SAFETY-32 round 5: stop closing sites, build an instrument that finds its own scope

Rounds 2, 3 and 4 each closed exactly the survivor they were handed and each time
the reviewer found the same shape one step over: the reduction, the call site, the
sibling argument, and now the sibling MODULE. Round 4's enumeration stopped at the
`Controller` class boundary; round 5's four blockers were reductions in validate.py,
over a list DERIVED from the roster three calls away from it. Closing four more sites
by hand would have produced a fifth rejection, so this round widens the instrument
instead and lets it enumerate.

⚠️ THE BLOCKERS ARE NOT EQUIVALENT MUTANTS - a safety refusal disappears, verified by
calling the shipped validator both ways on the real tree:

  roster max_w 5000 (inverter-a) beside 2000 (inverter-b), sign_disagree_w 3000
    clean                        -> 1 problem, "...max_w (inverter-b) is 2000 W, so the
                                    SAFETY-02 sign-disagreement watchdog can never fire"
    min(caps[:1])          +4 B  -> 0 problems. The site STARTS.
    caps_ix[:1]            +4 B  -> 0 problems.
  the SAME roster with the small cap FIRST (2000, 5000)
    clean                        -> 1 problem, naming inverter-a
    min(caps[-1:])         +5 B  -> 0 problems.

⚠️ AND THE ONE-ROSTER FIXTURE COULD NOT HAVE SEEN THAT. On [5000, 2000] the `[-1:]`
mutant still refuses, because last IS min there. A fixture of N members cannot separate
two reductions that agree on N members - which is why the behavioural half below is a
SET and not a roster.

THE INSTRUMENT (test_peakguard.py part 12j, widened past `Controller`)

Every `.py` under app/ is read off disk and parsed - NO MODULE IS NAMED IN THE CHECK -
and the roster is found by transitive closure from ONE seed, the `inverters` option key:
through returns, through lists ACCUMULATED in a loop over a roster (which is exactly
what `caps`, `roster` and `problems` are, and what round 4's rule could not express),
through attributes and properties, and THROUGH CALL ARGUMENTS, which is how it walks
into peakguard, arbiter, allocator, circulation, maintenance and store without being
told they exist. A slice of a roster stays a roster so a mutation cannot cut the taint;
an index is one unit and does not.

  183 roster reductions across 9 modules, out of 23 files parsed:
  main 96, allocator 24, validate 19, arbiter 14, circulation 10, maintenance 10,
  peakguard 4, inverter 3, store 3.  14 modules reduce over no roster at all.

⚠️ THE HAND-WRITTEN MODULE LIST WOULD HAVE BEEN WRONG IN BOTH DIRECTIONS. It named
watchdog.py, which reduces over no roster at all (SAFETY-30's sign watchdogs are
strictly per-unit), and omitted maintenance.py and store.py, which both do. A vacuous
claim about one module and no claim at all about two others - which is the argument for
computing the scope, made with data.

The pin rediscovers ALL FOUR blockers without being told where to look, and finds a
FIFTH beside them (validate.py:783, rule 5's own message) that no review round named.
Its counterfactual runs at two depths: every enumerated site's expression mutated with
each of the four cuts this card has been defeated by (183 x 4 = 732 spellings, all
reported), and `[:1]` written into app/validate.py at each blocker position with the
WHOLE analysis re-run over the mutated source.

THE SWEEP, AND THE NUMBER THAT MUST NOT BE QUOTED ALONE

All 183 sites mutated to `[:1]`, +4 B each, against the full 19-file suite in
goodwe-addon-ci (:ro, -B, fresh rsync per mutant, 0 .pyc before and after - a read-only
mount cannot write one): 183/183 RED. ⚠️ That is TAUTOLOGICAL on its own, because the
pin makes every slice red. So the 51 whose first red was test_peakguard.py were re-run
with the pin NEUTRALISED:

  145 of 183   a behavioural fixture separates the site from a slice of itself
   38 of 183   ONLY the structural claim. Behaviourally UNGRADED.

All 38 are listed by site with what a `[:1]` there would cost, in the DECISIONS record
and in part 12j itself: 15 are rows and counts on the operator's page, 3 are
`len(targets)` arithmetic beside a write path that IS graded, 6 are allocator internals
ALLOC-01 reaches but not with enough units to separate. NONE of the 38 is closed
behaviourally by this card and none is counted as covered.

THE BEHAVIOURAL HALF, because a shape pin grades an identity and never a meaning

test_validate.py grows a ROSTER SET for the three rules that reduce over
`unit_caps_indexed()`: widths 2, 3, 4, 5 with the one small cap at EVERY position,
14 rosters x 3 rules = 42 runs of the real validator, each asserting the refusal fires
AND names the unit the number came off. Beside it a COMPUTED family table of twelve
candidates - max, first, last, the mean, [:1], [-1:], [:2], [:3], [1:], [:-1], [::2] -
each compared against what the shipped validator actually answered on the same roster.
Only `min` agrees everywhere; a member no roster kills is printed as a NAMED SURVIVOR.

⚠️ A THIRD RULE WAS READING THE FLAT `max_w`, AND IT IS SAFETY-30's

Found by reading the three branches side by side, NOT by the pin - and said that way
because it is the shape of what the pin cannot see. `unit_sign_disagree_w` was graded
against `_number(opts, "max_w", 2000.0)`, three lines under a branch that had already
stopped doing that. Both directions were measured before the fix:

  two 2000 W units + a leftover `max_w: 5000` + unit_sign_disagree_w 3000 -> 0 problems.
  The site started, the status page said `armed`, and the per-inverter watchdog could
  not trip on EITHER unit. Identical sentence to the blocker, one branch over.
  two 5000 W units with NO flat key -> REFUSED against a 2000 W default the operator
  never wrote, for a threshold both units can reach.

It takes the same `weakest` the branch above takes and names the same unit. At a roster
of one `unit_caps()` is `[max_w]`, so the shipped single-inverter refusal is the same
comparison, the same boundary and the same sentence.

MUTATION PROOF (each: bytecode purged and counted 0 before and 0 after, the edit
re-read FROM DISK with the new text present and the old absent, byte delta printed)

  B1  validate min(caps)     -> min(caps[:1])        +4 B  RED test_validate.py
  B2  validate min(caps)     -> min(caps[-1:])       +5 B  RED test_validate.py
  B3  validate caps_ix       -> caps_ix[:1]          +4 B  RED test_validate.py
  B4  validate rule 6 min    -> min(caps_ix[-1:])    +5 B  RED test_validate.py
  B5  validate rule 5 msg    -> min(caps_ix[:1])     +4 B  RED test_validate.py
  B6  validate rule 5b msg   -> min(caps_ix[-1:])    +5 B  RED test_validate.py
  B7  CONTROL: weakest -> max(0.0, weakest)         +10 B  GREEN, as intended
  R3a peakguard max(floors)  -> min(floors)           0 B  RED test_peakguard.py
  R3b peakguard max(floors)  -> max(floors[-2:])     +5 B  RED test_peakguard.py
  R3c main call site packs   -> self.packs[:2]       +4 B  RED test_peakguard.py
  R4a main pack_floor_soc    -> vals[:2]             +4 B  RED test_peakguard.py
  R4b main pack_floor_soc    -> vals[-2:]            +5 B  RED test_peakguard.py
  R4c main pack_floor_soc    -> vals[:3]             +4 B  RED test_peakguard.py
  R4d main cycle() any(...)  -> self.inverters[:1]   +4 B  RED test_watchdog.py

B1-B6 are caught by the BEHAVIOURAL set in test_validate.py, before the pin is reached.
B7 is a deliberate no-op control: `max(0.0, weakest)` is `weakest` for weakest > 0, and
it stays green, so the battery is not merely red for everything.

R3a is a ZERO-byte delta, so the shipped function was called both ways as well:
  packs floors [15.0, 30.0, 20.0], peak_soc_floor 15
  binding_soc_floor(...) = 30.0   the 0-byte mutant would answer 15.0   (+15.0 pp)
  -> the site would discharge to 15 % while pack-b's own BMS cuts out at 30 %.

⚠️ PRODUCTION IS DEGENERATE ON THE FLOOR AXIS. SPIKE-04: pack A and pack B BOTH floor
at 10 %, so a differing floor exists ONLY in the fixtures. That is the argument for
keeping them, not for simplifying them - collapsing them deletes the only data in this
repository that can see the reduction at all. Restated in config.yaml's example, in the
probe and in the fixtures.

⚠️ INVERTER B DOES NOT PHYSICALLY EXIST (~2026-09-12). Every second-unit figure here -
4600 W FROM ITS DATASHEET, 15 % illustrative, the whole roster SET - is SIMULATED and
labelled. Only inverter A's 10 % is read off hardware.

REBASED onto 67e649a. Two conflicts with SPIKE-04a, both in the blocks this card had
rewritten (config.yaml's roster example and the probe's C/D sections). Resolved to carry
BOTH claims, and the probe was RE-RUN rather than hand-merged: `all cases passed` on
ghcr.io/home-assistant/amd64-hassio-supervisor:2026.08.0, transcript updated with the
date. MUTATION.md habit 8 - a committed transcript stitched together out of two runs is
not a measurement.

Suite: 19/19 green in goodwe-addon-ci, mounted :ro, bytecode purged and verified 0
before and after. CI-06: test_peakguard.py run four times in isolation, green each time
(288 ok lines); test_validate.py three times, green each time.

MUTATION.md gains habits 12 and 13: build the instrument once the same shape has been
found twice AND make it discover its own scope, and say what the instrument cannot see
with sizes in the same breath.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BwRLw7eKYj2byVdwB6kbjU

goodwe_controller/CHANGELOG.md

@@ -2,6 +2,544 @@
 
 ## Unreleased
 
+**Inverter B's rating is 4600 W, FROM ITS DATASHEET, and every documented copy
+of it now says so (SPIKE-04a).** Three places stated B's rating — `config.yaml`'s
+commented roster example, `DOCS.md`'s, and `tools/supervisor_shape_probe.py`'s
+"a blob an operator would POST" — and all three said **3000 W**, a number that
+was never measured, never read off a datasheet, and simply wrong. The owner
+supplied B's model and serial on 2026-09-04: a **GoodWe GW5048D-ES**, serial
+`55048ESU22AW0992`, max charging and discharging power **4600 W**. No control
+behaviour changes here; this is documentation and its pin.
+
+* **Serial characters 6-8 read `ESU`** → platform 105, AA55, the supported set.
+  Not `ESN`/`ESC` (ES G2, platform 745). SPIKE-01 does not gate this site.
+  ⚠️ A passed platform gate is **not** a proven control surface: `specs.md`
+  already calls other platform-105 serials *"same protocol family; untested
+  silicon"*, and this repo's `0x056E`/`0x056F` map was reverse-engineered from a
+  passive tap on **one BPS unit**. That check is the bench half of SPIKE-04.
+* **⚠️ 4600 W IS FROM A DATASHEET AND IS NOT A MEASUREMENT, and one ambiguity is
+  recorded rather than guessed.** The datasheet prints it under *Battery Input
+  Data*, so it is probably **DC**, while `max_w` is graded against `0x0518` —
+  the inverter's own **AC** power. If it is DC the AC ceiling is nearer
+  4600 × 0.94 ≈ **4324 W**, and the charge direction moves the other way
+  (≈ 4894 W drawn AC-side). **Neither reading is written in as fact.**
+  SAFETY-20's `min(max_w, dev_max)` already bounds it correctly under either,
+  once `dev_max` is read off the unit — so `max_w: 4600` is what to configure
+  today. `goodwe/NOTES.md` §1.1b carries the full record, with the date and the
+  method on every figure.
+* **A NEW CHECK, BECAUSE A DOCUMENTATION FIX WITHOUT ONE DRIFTS AGAIN.**
+  `test_validate.py` now pins all three documented rosters: they must be
+  findable, they must agree, and they must say **4600**. It also grades the
+  **provenance beside the number** — the datasheet it came from, the `0x0518`
+  it is graded against, and the 4324 W the DC reading would make it — **within
+  2500 characters of the roster, not file-wide.** The first version searched
+  whole files and was already vacuous on two of them, because `0x0518` appeared
+  in `config.yaml` and `DOCS.md` for unrelated reasons before this card: a
+  check whose data could not reach the failure it named, in the file that
+  exists to complain about those. Four mutations were run against the fixed
+  version — the probe back to 3000 (0-byte delta), the `DOCS.md` caveat
+  deleted, `config.yaml` alone "corrected" to 4324 (0-byte delta), and the
+  probe's entity renamed so the regex finds nothing — and all four are red.
+* **THE TEST FIXTURES KEEP 3000 W ON PURPOSE, AND NOW SAY WHY AT THE LINE.**
+  `test_inverter.py`, `test_maintenance.py`, `test_peakguard.py` and
+  `test_watchdog.py` build a second inverter at 3000 W against A's 5000 W. The
+  only property those checks consume is that B's number is **unmistakably** not
+  A's, and the real pair is a **400 W** gap — narrower than `slew_w`, inside the
+  range where an off-by-one-unit bug and ordinary control arithmetic produce
+  indistinguishable numbers. Collapsing them onto the real values would buy
+  realism those checks do not use and cost them their separating power, which
+  is the defect `test_peakguard.py`'s own §10g precondition exists to prevent.
+  They are **illustrative, not B's rating, and no longer a placeholder for it**.
+* **Two consequences with numbers, recorded in `goodwe/NOTES.md` §1.1b for the
+  cards that own them.** At 94 % battery-to-AC a round trip is ≈ 88 %, so A
+  charging 3 kW while B discharges 3 kW burns ≈ **360 W continuously** with the
+  meter at zero — SAFETY-31 argues its threshold from *"a few watts"* and has no
+  figure. And **< 13 W night self-consumption per unit** (≈ 26 W for two,
+  ≈ 0.6 kWh/day), which nothing in this add-on accounts for anywhere.
+* **The site is asymmetric in opposite directions:** A = 5000 W / 15.36 kWh
+  nominal, B = 4600 W / 20.48 kWh — ~33 % more energy, ~8 % less power. No
+  stability implication (ARCH-01's ḡ = Σαᵢgᵢ is a convex combination), but a
+  full recalibration of B moves 33 % more energy and **MAINT-05's 12 h `drain`
+  phase timeout was derived on A**; it should be re-checked against B rather
+  than assumed to carry.
+
+**Every pack is recalibrated on its own schedule, and only one moves power at a
+time (MAINT-05).** §12's monthly cycle drains a pack to empty and refills it to
+true its coulomb counter. There was exactly one `Maintenance` state machine and
+one un-namespaced history, so at two packs **only the first was ever
+recalibrated** and the second's counter drifted forever — and nothing prevented
+the cycle draining pack A while the loop or arbitrage charged pack B, which is
+two power-moving phases on one grid connection pulling against each other.
+
+* **ONE `Maintenance` PER ROSTER ENTRY**, each with its own schedule, phase and
+  history under `unit/<name>/...`. `store.migrate_maint_keys()` moves an
+  existing install's flat history into the first pack's namespace under
+  `migrate_unit_keys()`'s three rules — shared in code, not restated — with the
+  same read-back postcondition, and the flat keys are kept **live** as the
+  rollback view so a downgrade finds a current history rather than one dated
+  migration day.
+* **ONE MAINTENANCE SLOT PER SITE.** At most one pack may be in a power-moving
+  phase (`drain`, `charge` **and** `hold` — `charge` is in the set because a
+  pack pulled full from the grid beside a pack exporting to empty is 4 kW
+  circulating with the meter at zero). The holder is **derived** from the packs'
+  own stored phases, not persisted, so a restart mid-drain re-derives it. When
+  several packs are due in the same hour the slot goes to the one that has gone
+  **longest without a completed cycle**; roster order breaks a numeric tie and
+  nothing else. Nothing preempts a cycle in flight.
+* **THE DEFERRED PACK IS NOT STARVED AND IS NOT A FAULT.** A deferral does not
+  stamp `last_start_attempt` (that is the 90 min post-abort cooldown, inside a
+  60 min start hour), and does not move the phase, so neither the 12 h `drain`
+  nor the 36 h `charge` timeout — *"the battery is not following commands"* —
+  can be reached by waiting. It says WAITING, names the pack it waits for, and
+  recalibrates at the next start hour; a test drives the winner's whole cycle
+  and observes it start.
+* **THE SCHEDULE IS NOT STAGGERED BY CONSTRUCTION.** The interlock defers and
+  the stagger is what comes out of it: one contested month separates two packs
+  by one cycle length and they stay separated. `DECISIONS/MAINT-05-per-pack-maintenance.md`
+  §3 carries the argument.
+* **`validate.py` rule 6 now grades `maintenance_charge_w` against the WEAKEST
+  cap on the roster**, like rule 5. Its own comment said it would have to move
+  the day `maint_unit` stopped being `inverters[0]`; that day is this card. It
+  refuses configurations that used to start — a 3000 W charge beside a 2000 W
+  pack that is now recalibrated too — which is the `lo > hi` contradiction rule
+  6 exists to refuse, caught when the config is written. Identical at N=1.
+* **AT A ROSTER OF ONE, NOTHING CHANGES**, and it is asserted rather than
+  assumed: the state file keeps the flat keys and gains none, the migration does
+  not run, the four cycle events are the shipped strings character for
+  character, and the status banner is the bare phase with no pack name.
+* **THE MIGRATION IS GRADED AGAINST THE ROSTER, NOT AGAINST THE NAME IT WAS
+  HANDED** (fix round). `migrate_maint_keys()` takes the roster rather than a
+  pack count and `_verify_moves()` checks the namespace it wrote against the
+  names the machines actually read. A postcondition that reads back the
+  namespace it was given cannot detect being given the wrong one: handed
+  `inverter.UNIT_A` on a roster of `["garage", "shed"]`, the first version of
+  this card wrote all four keys under `unit/inverter-a/...`, read all four back
+  equal, reported **verified**, and stamped `maint_schema_version` — so both
+  packs came up believing they had never been recalibrated and the next start
+  skipped the migration too. That is now `misdirected`: not `ok`, the marker is
+  **not** written, and the next start re-runs and recovers (rule 1 makes the
+  re-run a no-op). `build_roster()` numbers a nameless roster entry
+  `inverter-1`, so `inverter-a` is not even the fallback for a declared roster.
+  Every two-pack fixture that could see this now names its first pack
+  `garage`; with them all called `inverter-a`, `self.inverters[0].name` and
+  `UNIT_A` were the same string and the swap left all 19 suites green.
+* **AND THE REFUSAL NOW SURVIVES THE NEXT START** (second fix round). The fix
+  above graded only the keys it had just *written*, so it refused exactly once:
+  measured at Controller level over one persisted store, roster
+  `["garage", "shed"]`, with the call site still wrong on both starts —
+  construction 1 `ok=False misdirected=4 moved=4 kept=0 marker=<absent>`,
+  construction 2 `ok=True misdirected=0 moved=0 kept=4` and the marker
+  **stamped**. On the second start rule 1 finds the wrong keys already there and
+  routes all four to `kept`, so `moved` is empty and the roster loop grades
+  nothing — and start 3 skips the migration forever, *including after the call
+  site is fixed*. Loud-and-recoverable became silent-and-permanent after one
+  restart. `_verify_moves()` now grades `set(moved) | set(kept)`; both lists are
+  built from the same `unit` argument, so grading `kept` grades the same
+  namespace decision. `test_maintenance.py` part **9c** drives two starts with
+  the same wrong unit over one store — the shape nothing in the suite had, since
+  9b measured recovery along the path where the operator has *already* fixed the
+  call site.
+* **AND THE MARKER GATE IS PINNED ON BOTH GUARDS.** `if rep.ok:` is
+  `not lost and not misdirected`, and the mutation `if rep.ok or rep.kept:`
+  stamped the marker on a migration that **lost a month of history** with all 19
+  suites green. Part **9d** is the fixture that separates the two guards: two
+  keys already namespaced (`kept` non-empty) and two whose write the store
+  drops (`lost` non-empty), `misdirected` empty throughout.
+* ⚠️ **AND THE SAME MUTATION ONE FUNCTION UP WAS A SURVIVOR, FOUND HERE AND
+  CLOSED.** `migrate_unit_keys()` — SAFETY-02's **latch** migration — gates its
+  own marker on `rep.ok` too, and every check that reached that gate ran on a
+  store where nothing was namespaced yet, so `rep.kept` was empty in all of
+  them. `if rep.ok or rep.kept:` there stamps `schema_version` on a migration
+  that **dropped the latch**: the guard is off, the file still says
+  `sign_tripped: true`, and the next start skips the migration that would have
+  re-armed it. Measured a survivor of all 19 files; now red on
+  `test_inverter.py` part 10, which drops the `sign_tripped` write beside an
+  already-namespaced `sign_reason`. No source change — the shipped gate was
+  right; nothing could see it.
+* ⚠️ **FOUND, RECORDED, NOT FIXED (SECOND ONE).** `migrate_unit_keys()` —
+  SAFETY-02's latch migration — has the identical hole and it costs a **guard**
+  rather than a history. It runs before `build_roster()` (each `Inverter`
+  restores its latch in its constructor), so it is handed the constant `UNIT_A`
+  and there is no roster to grade against. On a declared roster a pre-DEV-02
+  `sign_tripped: true` lands under `unit/inverter-a/sign_tripped`, which no unit
+  reads — and `_grade_store_migration()`, whose job is to re-arm exactly this,
+  is **defeated by the key the defect wrote**: its "is anything namespaced
+  latched" test is satisfied by the misdirected value itself. Fixing it means
+  naming the migration's target from the config before any `Inverter` exists,
+  which is `inverter.py` and SAFETY-02's semantics, with its own battery.
+  Measured and pinned as a running check (`test_maintenance.py` part 13), whose
+  assertions state the WRONG behaviour on purpose.
+* ⚠️ **FOUND, RECORDED, NOT FIXED.** A maintenance drain **pins** its pack while
+  the loop's target is a **site** claim, so when the site target is below the
+  pin the only split asks the sibling to charge against the drain and
+  `allocator.plan()` refuses it — the site parks at 0 W and the pack's 12 h
+  timeout would eventually report *"drain stuck"* about a healthy battery.
+  Pre-existing on `release/1.0` at `2d025ee` with the unit names swapped; fixing
+  it means changing what the loop's target means while a unit is pinned, which
+  is loop and allocator territory. Pinned as a running check
+  (`test_maintenance.py` part 12) so it cannot be forgotten.
+
+**`bms_soc_floor` describes a BATTERY, so at two packs it stops being one number
+(SAFETY-32).** The peak guard's energy backstop rests on two figures: the EMS
+floor `peak_soc_floor` (the depth a money-saving discharge may never reach) and
+`bms_soc_floor`, the percent **that pack's own BMS** cuts out at — register
+`0x0560`, read as 10 on the reference GW5000S-BP. Rule 8 refuses an EMS floor at
+or below the cut-out, because such a floor can never bind: the BMS wins and the
+status page reports a backstop that does not exist. With ONE cut-out for two
+packs of different models that rule graded the EMS floor against whichever
+battery happened to be written into the flat key, and a site with cut-outs of
+10 % and 30 % under `peak_soc_floor: 20` **started** — the guard would then
+command a forced discharge that pack B's BMS, not the operator's floor, is the
+thing that stops. Rule 8's own failure mode, reached through rule 8.
+
+* **`bms_soc_floor` becomes a per-entry roster key, and it is REQUIRED there.**
+  Nothing is inherited: an entry that omits it has no cut-out, does not borrow
+  the flat one, and is refused by name at startup. A per-entry *default* was
+  considered and declined — a default is one number for every pack, which is the
+  defect this card removes; `DECISIONS/SAFETY-32-per-pack-bms-floor.md` records
+  the decision and its migration cost. The flat key is unchanged and is now
+  documented as the single-inverter shape's own cut-out.
+* **Rule 8 grades EVERY unit and NAMES the one it refuses**, through a new
+  `unit_bms_floors_indexed()` that carries the **roster** index the value came
+  off. CFG-04 had just fixed a rule 5/6 bug of exactly this shape — a caps-list
+  index fed to a function that indexes the roster, so the message named the wrong
+  inverter — and this reader skips nothing, so there is no second list for an
+  index to desynchronise from. Graded on a roster whose offender is the SECOND
+  entry *and* on the same two numbers in the other order.
+* **The permission is answered PER PACK.** `Controller.pack_floor_soc` — the
+  emptiest measured pack — is untouched and is still what `decide()` grades;
+  what changed is the floor. `peakguard.pack_permissions()` answers one
+  `(name, reason)` per pack against that pack's own floor, and the floor-limited
+  notification now names the pack that stopped the site rather than "the
+  battery". At a roster of one the alarm is byte-for-byte what SAFETY-25 shipped.
+* **An unstated cut-out withholds rather than defaults.** `None`, not `0.0` —
+  `discharge_permitted()` already treats an unknown floor as no permission, so
+  the absence costs a shave and never a pack, including on `sim/inject.py`'s path
+  which does not run the startup gate.
+* ⚠️ **Stated rather than hidden:** under any config rule 8 accepts, the per-pack
+  *floor* reduces to `peak_soc_floor` for every pack — rule 8 makes it so. The
+  reduction is written for `unit_caps_indexed()`'s reason ("unreachable" is a
+  claim about rule ordering, not a property of a function) and is graded on both
+  sides: 1775 rule-8-valid `(peak_soc_floor, bms_soc_floor)` pairs answer the
+  option, and the pair rule 8 refuses answers the cut-out. `peak_soc_floor`
+  itself is still one number for the site, so the "deeper pack needlessly
+  withheld" half is filed as an open question, not closed.
+* **`tools/supervisor_shape_probe.py` re-run and repaired.** MUTATION.md habit 8:
+  a committed transcript is a check and it decays. The required key broke §C and
+  left §D printing PASS on a refusal about a *different* key. §D is now an
+  isolated counterfactual over an accepted, merged baseline, one key varied per
+  case, asserting the key **name** in the refusal — `all cases passed` on
+  `amd64-hassio-supervisor:2026.08.0`.
+* ⚠️ **SECOND FIX ROUND — the round-2 fix carried the defect it was written to
+  close, in its prose.** Two more rewrites survived a green 19-file suite on
+  `8501e05`: `binding_soc_floor()`'s `max(floors)` → `max(floors[-2:])` (+5
+  bytes) and the call site's `self.packs` → `self.packs[:2]` (+4 bytes). Both are
+  checks whose TEXT claimed more than their FIXTURE could reach — 12c2's only
+  counterfactual varied the **last** pack, which every window ending at the
+  roster's end follows, and all three 12g runs drove a **two**-pack roster, where
+  `[:2]` *is* the roster. **The general law, now stated in the file: a fixture of
+  N packs cannot separate two reductions that agree on N packs**, and the regress
+  does not terminate in a fixture — `[:1]` needed two packs, `[:2]` three, `[:3]`
+  four. So the reduction is now graded by a **length sweep** (78 simulated
+  rosters, lengths 1–12, deepest cut-out at every position, closing head and tail
+  windows of width 1–11 at once) plus a **computed family table** of fourteen
+  rewrites that goes red naming any member no roster kills; the call site is
+  graded by **three-pack runs** with the deep cut-out at index 0, 1 and 2, plus
+  an **exact structural pin** (the roster argument must unparse to `self.packs`)
+  for the slice widths no bounded roster can reach. Round 3's own battery found
+  `self.packs[:3]` / `[-3:]` surviving all six behavioural runs — the structural
+  pin is what closes them — and one row is recorded **NOT CAUGHT**: a reduction
+  over a window twelve packs wide, which needs a 13-pack roster to observe.
+  No shipped behaviour changed in this round; only what observes it.
+* **Twelve source mutations, eleven red, one argued survivor**, through
+  `tools/mutation_run.sh` (bytecode purged). The battery also found a real
+  weakness of its own: `_flat_bms_floor()` had the default in two places and each
+  covered for the other, so deleting either survived — masking, in three lines of
+  option reading. One defaulting line now.
+* ⚠️ **FIX ROUND — that battery was INCOMPLETE, and the review found what it
+  missed.** Neither of the two mutations below appears in the twelve rows above:
+  the ledger graded "the reduction dropped" and "the bare option handed over"
+  and never asked what happens when the reduction is simply the *wrong* one, or
+  when the roster handed to it is *sliced*. No shipped behaviour changed here;
+  two *checks* whose data could not reach the failure they named did. `max(floors)` → `min(floors)` in
+  `binding_soc_floor()` (zero byte delta) and `self.packs` → `self.packs[:1]` at
+  the `decide()` call site (+4 bytes) were each green across all 19 files. Both
+  survived for the same reason: **under any rule-8-valid config every pack's
+  `pack_floor_pct()` equals `peak_soc_floor`**, so `floors` is a list of
+  identical numbers and `max`, `min`, first and last are four spellings of one
+  answer — and the one check using a cut-out above the EMS floor ran on a roster
+  of ONE, where every reduction agrees. The second also walked through part 12g,
+  which asserted `"self.packs" in <the unparsed call site>`: a slice still
+  contains that substring, so the pin passed on the mutation that grades the
+  whole site against pack A. MUTATION.md habit 4 — a source-text test is a check
+  on how code is spelled, not on what it does.
+  * **New part 12c2** evaluates `binding_soc_floor()` on a roster whose per-pack
+    floors **all differ** — four simulated cut-outs of 30/25/40/35 %, every one
+    above `peak_soc_floor`, chosen so that `max` (40), `min` (25), first (30),
+    last (35) and the mean (32.5) are five different numbers and the two extremes
+    sit at *interior* roster positions. Two packs cannot separate those four;
+    four is the smallest roster that can. The precondition is its own check
+    (habit 1) and one pack — the last, neither deepest nor shallowest — is then
+    raised past the old maximum and the answer follows it (habit 7).
+  * **Part 12g is now behaviour, not source text.** Real `Controller.cycle()`s on
+    a two-pack roster, both packs at the same SoC, with the cut-outs as the only
+    thing varied: move the **non-first** pack's cut-out and the floor reaching
+    `decide()` moves 20 % → 35 %, with the site's verdict flipping from a sized
+    discharge to FLOOR-LIMITED. `self.packs[:1]` answers 20 % both times. Which
+    option key the floor comes from stays pinned on the source where it already
+    was, because that claim genuinely is about the code's shape.
+  * ⚠️ **Two MORE survivors, found by this round's own battery and recorded
+    rather than quietly repaired.** (a) The first cut of part 12g varied only
+    pack B's cut-out, so `self.packs[-1:]` — grade the site against the **last**
+    pack — agreed with the full roster on both runs and stayed green across all
+    19 files (+5 bytes). A two-run fixture pins one slice and not the other.
+    Part 12g now runs the **same pair of cut-outs in both orders**, so the
+    deeper one is on the second pack in one run and on the first in another and
+    the site floor is 35 % in both; no single-position slice passes all three
+    runs. (b) `binding_soc_floor()`'s empty-roster branch rewritten to
+    `return None` (−43 bytes) was also green — a defensive branch nothing in the
+    suite reached. Both directions of it are pinned now, with the note that it
+    cannot permit anything either way because `pack_floor_soc` is `None` at an
+    empty roster.
+
+* ⚠️ **THIRD FIX ROUND — the same family was alive at the SIBLING ARGUMENT of
+  the same call.** Rounds 2 and 3 hardened the FLOOR axis; `decide()` is handed
+  a second reduction over the same roster beside it, `soc=self.pack_floor_soc`,
+  and `peakguard.py`'s own docstring names the two as a pair. Three rewrites of
+  `Controller.pack_floor_soc` were green across all 19 files on `52380c0` and
+  again after the rebase onto `3934ad7`: `min(float(v) for v in vals)` →
+  `vals[:2]` (+4), `vals[-2:]` (+5) and `vals[:3]` (+4), while `[:1]`, `[-1:]`,
+  `[1:]`, `[:-1]`, `max()` and the mean were all red. **The suite's separating
+  power on the SoC axis stopped at exactly N=2** — the state the floor axis was
+  in before round 3. It changes a real answer: a three-pack roster at
+  50 / 60 / 11 % with the emptiest pack THIRD answers 11 % shipped and 50 %
+  mutated, so the site is graded for discharge PERMISSION against a pack 39
+  points fuller than the emptiest and spends a pack its own BMS stops. Round 3's
+  machinery transfers directly — new **part 12i**: six real `cycle()`s with the
+  emptiest pack at every position of a three-pack roster and both positions of a
+  two-pack one (the observable read off `PeakDecision.soc`, cut-outs held flat
+  so the floor axis cannot be what moved the answers); computed slice and
+  non-slice tables that go red naming any member nothing kills; a **length
+  sweep** of 78 real rosters, lengths 1–12, closing head and tail windows of
+  width 1–11 and recording width 12 as honestly unreached; a **none-guard
+  sweep** over the same 78; and **three exact structural pins** — the `vals`
+  comprehension must iterate the bare `self.inverters`, and both `any()` and
+  `min()` must reduce the whole `vals` — which close every slice width at every
+  roster length. No shipped behaviour changed; only what observes it.
+* ⚠️ **SPIKE-04: production is DEGENERATE on the floor axis, so the
+  differing-floor fixtures are the entire exposure.** Pack B is 4 × Dyness
+  DL5.0C, DoD 90 → floor 10 % — **FROM THE DATASHEET AND THE SERIAL; inverter B
+  does not exist and nothing about it is measured.** Pack A is DoD 90 → 10 % as
+  well, and that one *is* measured. Both packs therefore floor at the same value
+  in the field, `max(10, 10)` is 10 whatever it reduces, and a differing floor
+  exists **only in the fixtures**. That is the argument for keeping them, not for
+  simplifying them: anyone who collapses `A_PACK_FLOOR != B_PACK_FLOOR_SIM` to
+  one number deletes the only data this repository has that can see the
+  reduction at all, and part 12c2's precondition check goes red if they do. The
+  **SoC** axis has no such degeneracy — two batteries at the same state of
+  charge is a coincidence, not a configuration — which is what makes
+  `pack_floor_soc` the half of the pair that will be live from the day inverter
+  B is wired up.
+* ⚠️ **The FAMILY enumerated rather than the member closed, and it found four
+  more survivors in code this card never touched.** Every reduction over
+  `self.inverters` / `self.packs` in `Controller` was found from the AST and
+  mutated to a slice against the full suite. Eighteen were already graded
+  (`site_batt_w`, `site_rating_w`, `site_max_w`, `site_dev_bound`,
+  `site_circulating_w`, `inputs_ok`, `units`, `Controller.packs`, `maint_unit`,
+  the maintenance-due minimum, the store-migration roster, `validate_unit_ranges`
+  and the round-2/3 sites). **Four were not, and three of them are SAFETY-30's,
+  all `any(i.sign_wd.tripped for i in self.inverters)`** — `cycle()`'s
+  stale-input hold, `publish()`'s status vocabulary and `status_payload()`'s
+  banner — each +4 bytes and green across all 19 files, for the reason every
+  blocker on this card was invisible: **the only fixture that reached them was a
+  roster of ONE, where `self.inverters[:1]` IS `self.inverters`.** The first has
+  teeth: a latched unit must take the roster past that hold or its `failsafe`
+  SET stops being re-filed the moment a poll is missed, which is main.py's own
+  sentence for it. New **test_watchdog.py part 8** grades all three with the
+  latch at EACH position plus a no-latch counterfactual, and new **part 12j**
+  pins all 26 roster comprehensions in `Controller` to iterating the roster
+  whole — so a reduction added over a slice anywhere in the class is red on the
+  line it is written, with no fixture required. Two rows are reported open
+  rather than closed: `_grade_store_migration()`'s second disjunct (now red on
+  12j, and argued to be an equivalent mutant because a unit tripped at startup
+  implies its namespaced key is set) and `cycle()`'s undeliverable-unit **log
+  line**, where `sorted(...)[:1]` (+4 bytes) is still green — it truncates names
+  in one event message; the refusal itself is a different expression and is
+  graded, so the site still stops.
+* ⚠️ **FOURTH FIX ROUND — the same family one MODULE over, and the pin widened
+  until it discovers its own scope.** Round 4 enumerated the family and closed
+  it inside the `Controller` class; the review found four more survivors
+  immediately, in **validate.py**, on a 19/19 green suite: `min(caps)` →
+  `min(caps[:1])` (+4) and `→ min(caps[-1:])` (+5) at `:779`, the caps
+  comprehension over `caps_ix[:1]` (+4) at `:778`, and
+  `min(caps_ix[-1:], key=…)` (+5) at `:893`. **None is an equivalent mutant.**
+  With `max_w: 5000` on inverter-a beside `2000` on inverter-b and
+  `sign_disagree_w: 3000`, the shipped validator refuses — *"the SAFETY-02
+  sign-disagreement watchdog can never fire"* — and all four **accept**: the
+  site starts with a watchdog reporting `armed` that cannot trip on pack B.
+  Rule 5's own failure mode reached through rule 5; rule 6 loses its refusal the
+  same way. **No shipped behaviour changed by the pin; what changed is that a
+  hand-written scope stopped being trusted.**
+  * **`test_peakguard.py` part 12j now parses every `.py` under `app/`** — no
+    module is named in the check — and finds the roster by transitive closure
+    from **one seed**, the `inverters` option key: through returns, through
+    lists **accumulated in a loop over a roster** (which is what `caps`,
+    `roster` and `problems` are, and what round 4's rule could not express),
+    through attributes and properties, and **through call arguments**, which is
+    how it walks into `peakguard`, `arbiter`, `allocator`, `circulation`,
+    `maintenance` and `store` without being told they exist. A slice of a roster
+    stays a roster so a mutation cannot cut the taint; an index is one unit and
+    does not. **183 reductions across 9 modules** out of 23 files parsed — `main.py` 96,
+    `allocator.py` 24, `validate.py` 19, `arbiter.py` 14, `circulation.py` 10,
+    `maintenance.py` 10, `peakguard.py` 4, `inverter.py` 3, `store.py` 3.
+  * **The hand-written module list would have been wrong in BOTH directions.**
+    It named `watchdog.py`, which reduces over no roster at all — SAFETY-30's
+    sign watchdogs are strictly per-unit — and omitted `maintenance.py` and
+    `store.py`, which both do. A vacuous claim about one module and no claim
+    about two others.
+  * **The pin rediscovers all four blockers without being told where to look,**
+    and finds a fifth beside them (`validate.py:783`, rule 5's message) that no
+    review round named. Its counterfactual runs at two depths: every enumerated
+    site's own expression mutated with each of the four cuts this card has been
+    defeated by (732 spellings, all reported), and `[:1]` written into
+    `app/validate.py` at each blocker position with the **whole** analysis
+    re-run over the mutated source — a sliced site required at that exact line,
+    where the clean package reports none.
+  * **And the behavioural half, because a shape pin grades an identity and never
+    a meaning.** `test_validate.py` grows a **roster SET** for the three rules
+    that reduce over `unit_caps_indexed()` — widths 2, 3, 4, 5 with the one
+    small cap at every position, 42 runs of the real validator — plus a
+    **computed family table** of twelve candidate reductions each compared with
+    what the shipped validator actually said on the same roster. Only `min`
+    agrees everywhere; a member no roster kills is printed as a named survivor.
+  * ⚠️ **A THIRD RULE WAS READING THE FLAT `max_w`, AND IT IS SAFETY-30's** —
+    found by reading, **not** by the pin, which is the shape of what the pin
+    cannot see. `unit_sign_disagree_w` was graded against
+    `_number(opts, "max_w", 2000.0)` three lines under a branch that had already
+    stopped doing that. **Both directions were measured:** two 2000 W units
+    beside a leftover `max_w: 5000` and `unit_sign_disagree_w: 3000` produced
+    **zero problems** — the site started with a per-unit watchdog that could not
+    trip on either unit — and two 5000 W units with no flat key were **refused**
+    against a 2000 W default the operator never wrote. It takes the weakest
+    unit's own cap now and names that unit; at a roster of one it is the same
+    comparison, boundary and sentence it always was.
+  * ⚠️ **What the pin cannot see, with sizes**, in
+    `DECISIONS/SAFETY-32-per-pack-bms-floor.md` §4d: semantics (`min` vs `max`
+    is a zero-byte edit); direct index sites that are not reductions; a
+    subscript by a *variable* (`weights[u]` is why); a slice through an
+    intermediate call; anything outside `app/`, including `sim/inject.py`; and
+    the taint's own seam at `*args` and unresolvable receivers, where everything
+    downstream goes **unenumerated rather than reported unreachable**.
+    `cycle()`'s undeliverable-unit log line is still green at +4 bytes and still
+    filed open, unchanged from round 4.
+  * **`tools/supervisor_shape_probe.py` re-run again (2026-09-05, 2026.08.0).**
+    The rebase onto `67e649a` merged SPIKE-04a's `max_w: 4600` into the same
+    §C/§D blocks this card had rewritten, and a transcript stitched by hand out
+    of two runs is not a measurement. `all cases passed`, exit 0, with B's
+    rating reading 4600 — its **datasheet** figure, still not a measurement of
+    any unit.
+  * ⚠️ **THE SWEEP, AND THE NUMBER THAT MUST NOT BE QUOTED ALONE.** All 183
+    sites mutated to `[:1]`, one at a time, against the full 19-file suite in
+    `goodwe-addon-ci` (`:ro`, `-B`, fresh rsync per mutant, 0 `.pyc` before and
+    after): **183/183 RED** — which is tautological on its own, because the pin
+    makes every slice red. The 51 whose first red was `test_peakguard.py` were
+    re-run with the pin **neutralised**: **145 of 183 have a behavioural fixture
+    that separates them from a slice of themselves; 38 have only the structural
+    claim.** All 38 are listed by site, with what a `[:1]` there would cost, in
+    §4d and in `test_peakguard.py` part 12j — 15 of them are rows and counts on
+    the operator's page, 3 are `len(targets)` arithmetic beside a write path
+    that IS graded, 6 are allocator internals ALLOC-01's fixtures reach but not
+    with enough units to separate. **None of the 38 is closed behaviourally by
+    this card, and none is counted as covered.**
+
+⚠️ **No measurement of inverter B appears anywhere in this change.** It does not
+exist yet. Every second-pack figure — in fixtures, in the probe, in
+config.yaml's commented example — is **simulated and labelled**; only inverter
+A's 10 % is read off hardware.
+
+**At two inverters the sign watchdog stops comparing one unit against the whole
+site meter, and becomes two checks instead of one (SAFETY-30).**
+`disagrees(commanded_w, grid_w, ...)` compared the command **this controller put
+on ONE actuator** against **what the whole site** is exchanging with the network.
+At a roster of one those are two views of the same thing - §6.2's identity - and
+that is what makes the check sound and what earned it the best record in the
+safety table. At two they are not, and it fails **in both directions**: inverter
+A commanded −1500 W while the meter reads +1500 W is evidence of nothing when
+inverter B is discharging 3000 W exactly as told, and - worse - a unit genuinely
+driving the wrong way is **MASKED** by a sibling that covers for it, because the
+site figure nets. No threshold recovers that; the summation destroys the
+evidence.
+
+* **PER UNIT: each inverter's own share against its OWN measured power**
+  (`Inverter.batt`, holding register 0x0518, per-instance since DEV-02).
+  Conclusion: *the actuator I am driving is not doing what I told it* - evidence
+  about one device, so the remedy stays the per-unit `failsafe` SET that
+  `arbiter.CLAIM_SCOPE` has classified `SCOPE_UNIT` since it was written. That
+  entry read *"the evidence is site-wide and the remedy is not"*; after this card
+  the first half is simply true. Verified by making one unit reverse while its
+  sibling covers it at the meter: the shipped comparison is false on **0 of 14**
+  cycles while the new one convicts that unit on every one, and only that unit is
+  parked.
+* **SITE: the SUM of the standing commands against the meter**, as a new
+  `sign_site` SET at `P_SAFETY` on the site actuator, latched under flat
+  `site_sign_tripped` / `site_sign_reason`. This is the fault the 2026-08-20
+  capture contains and **no per-unit check can see it**, because every inverter
+  is faithfully executing its share - what is wrong is the number they were
+  given. It stops every inverter and the banner names the site rather than an
+  inverter, so nobody is sent to swap the wrong box.
+* **⚠️ Different questions from SAFETY-31, and the PR and the decision record
+  both say which.** `circulation.py` compares the units **against each other**
+  and catches two inverters pushing *through* each other, which nets to zero at
+  the meter and which no single command is evidence about. This compares each
+  unit against **its own command**. They overlap in exactly one scenario - a unit
+  reversing while its sibling pulls the other way - and `test_watchdog.py`
+  asserts on that fixture that both fire and what each concludes. It is also not
+  the saturation freeze: that is a magnitude test (`|command − measured| >
+  500 W`) that shapes the loop and fires on a merely saturated unit; this is a
+  direction test that latches and parks one.
+* **⚠️ All four thresholds RE-DERIVED at two inverters - one carried with a
+  justification, one refused.** 1890 closed-loop runs in which nothing is faulty,
+  over the whole range `config.yaml` accepts for `gain` and `slew_w`, across load
+  reversals with write staggers and flapping loads, **run at both rosters so the
+  N=1 arm is the control.** `unit_sign_disagree_w: 250` is `saturation_w / 2`, so
+  the smallest divergence this can see is exactly the gap at which the control
+  law itself already says the inverter is not following; `unit_sign_disagree_s:
+  85` is 8.5× the worst legitimate opposition those runs produce (10.0 s), 2.8×
+  the firmware's own 30 s link veto and under half the site hold.
+  `site_sign_disagree_w: 300` is **carried** - its argument is the meter's noise
+  floor and the deadband, and the roster touches neither. **`site_sign_disagree_s`
+  is NOT: 180 s would have shipped a false trip.** The worst legitimate crossing
+  at 300 W is 110 s with one inverter and **215 s with two**, because the command
+  must travel a rail that is the roster sum at a `slew_w` that is not - reachable
+  at `gain: 0.05`, `slew_w: 50` and a 2.2 kW load reversal, all schema-legal.
+  325 s is 1.51× that, inside the 900 s billing quarter and inside every episode
+  of the fault day. See `DECISIONS/SAFETY-30-per-unit-sign.md`.
+* **⚠️ At ONE inverter this changes nothing, and "nothing" is asserted rather
+  than assumed.** The two questions are the same sentence at a roster of one, so
+  only the detector that shipped is armed: same comparison against the meter,
+  same 300 W / 180 s, same reason string byte for byte, same namespaced latch and
+  flat rollback mirror, same `failsafe` claim, same banner, same single status
+  row - and the site detector's window is never so much as opened.
+  `unit_sign_disagree_*` and `site_sign_disagree_*` do nothing on a
+  one-inverter install.
+* **The threshold WIRING is pinned, not just the threshold values.** Review
+  found that `arm(self.sign_cfg if one else self.unit_sign_cfg, ...)` could be
+  collapsed to `arm(self.sign_cfg, ...)` — arming every unit with the
+  meter-facing pair instead of the re-derived per-unit one — with all 19 test
+  files still green, because nothing at a roster of two read `sign_wd.cfg` and
+  the trips in the fixture happen under either pair. `test_watchdog.py` part 3
+  now asserts the pair and the instrument on **every** detector against
+  literals, and the fixture's three threshold pairs were made pairwise distinct
+  in both coordinates so that a hold delivered from the wrong pair cannot
+  compare equal to the right one. M14–M16 in
+  `DECISIONS/SAFETY-30-per-unit-sign.md` §5.
+* **Known and recorded rather than fixed:** the site comparison is defeated by a
+  square-wave load whose half period is about one control cycle, which holds it
+  opposed for ~1000 s at *both* rosters. The N=1 control arm is what says this is
+  a property of comparing a lagged command against a meter and has nothing to do
+  with the second inverter, so it is not acted on here - it wants an alternation
+  test, not a number, and a card of its own.
+
 **An operator can declare a second inverter, and the shape was measured against
 the Supervisor rather than assumed (CFG-04).** DEV-02 made rating, SoC,
 watchdogs, heartbeat and store keys per-instance, so the code no longer assumed

goodwe_controller/DECISIONS/MAINT-05-per-pack-maintenance.md

@@ -0,0 +1,388 @@
+# MAINT-05 — one recalibration per pack, one at a time, and what the card left open
+
+**Status: DECIDED 2026-09-04.** There is one `Maintenance` per roster entry,
+each with its own schedule, phase and history. The site holds exactly **one
+maintenance slot**: at most one pack may be in a power-moving phase, the slot is
+derived from the packs' own stored phases rather than persisted separately, and
+when several packs are due in the same hour it goes to the one that has gone
+**longest without a completed cycle**. The schedule is **not** staggered by
+construction — the interlock defers, and the stagger emerges from it.
+
+**Fix round 2026-09-04.** The store migration is now graded against the
+**roster** rather than against the unit name it was handed — see §2's
+*"The read-back postcondition is graded against the ROSTER"* — and the same
+hole in SAFETY-02's latch migration is recorded in §7.
+
+Filed with MAINT-05. Card `6a9a9e8503872905f298d80a`.
+
+---
+
+## 0. What the card asked to be decided, and the answers
+
+| the card's open question | the decision |
+|---|---|
+| §2 the arbitration | **one pack at a time; the other stays under the loop.** Nothing preempts a cycle in flight. |
+| §2 who wins when several are due | **the longest-waiting pack**, `last_completed` oldest first, `inf` for never-recalibrated. Roster order breaks a numeric tie and nothing else. |
+| §4 staggering | **not by construction.** The interlock defers the second pack; the stagger is what comes out of it. §3 below. |
+| AC 7 rule 6's `maint_unit` comment | **no, the first entry's cap is no longer right.** It grades against the weakest cap on the roster now, like rule 5. §4 below. |
+| the store layout at N=1 | **flat, unchanged.** The namespace appears the day the roster grows. §2 below. |
+
+## 1. Why the slot is derived and not persisted
+
+The holder is *"whichever pack's stored phase is not `idle`"* — a fact already on
+disk, written by the packs themselves in the per-pack history this card
+introduces. A `maint_slot_holder` key would be a **second opinion about the same
+thing**, and the two would disagree the first time the add-on was killed between
+writing the phase and writing the holder. Derived, a controller that comes back
+up mid-drain reads the drain and re-derives the holder from it; there is no
+recovery path to get wrong because there is no second fact.
+
+`CHARGE` is in `POWER_MOVING` even though the control **loop** owns the setpoint
+in that phase, and that is the decision most likely to be "simplified" later.
+Pack A charging from the grid at 2000 W while pack B drains 2000 W to export is
+4 kW going round in a circle with the meter reading zero and both packs
+believing they are behaving — the circulating power [[SAFETY-31]] exists to
+detect, arriving as something this module can simply not do. `HOLD` is in the
+tuple for the other half of the sentence: it moves nothing, and it **owns the
+actuator**, so a second cycle starting under it would still be two
+recalibrations at once.
+
+## 2. Why the store layout at a roster of one does not change
+
+The maintenance keys are flat — `phase`, `phase_started`, `last_completed`,
+`last_start_attempt` — on every install since the add-on shipped. At one pack
+there is nothing for a namespace to distinguish that pack **from**: §6.2's
+identity, the same one `arbiter.actuator_for()` leans on when it routes every
+claim to the site actuator at N=1. So `Maintenance` is constructed with
+`unit=None` there, reads and writes the flat keys directly, and
+`migrate_maint_keys()` **returns without writing anything**. A one-inverter state
+file comes out of this card byte-identical, which is AC 6 and which
+`test_maintenance.py` part 10 asserts by driving a whole cycle and comparing the
+key set.
+
+### The migration has its own marker, and that is not tidiness
+
+`SCHEMA_VERSION` / `schema_version` is DEV-02's, and it is **monotone**: every
+install runs that migration exactly once. This one runs **only where the roster
+has more than one pack**, so a single counter could not express *"migrated on
+this site, deliberately not migrated on that one"* without lying to the next
+version that reads it — a one-inverter store would have to be pinned at the old
+number forever, and a genuine future v2 migration could then never be told apart
+from a site that simply never grew. `maint_schema_version` answers one question
+and answers it honestly.
+
+The **three rules are shared in code**, not restated: `_move_flat_keys()` is
+rule 1 (never overwrite an existing namespaced value) and rule 2 (the flat keys
+stay in the file), `_verify_moves()` is the read-back postcondition, and the
+marker is written last. A second transcription of a migration is a second
+opinion about what "it landed" means.
+
+### The read-back postcondition is graded against the ROSTER (fix round)
+
+**A postcondition that reads back the namespace it was handed cannot detect
+being handed the wrong one.** That sentence is the whole finding, and this card
+shipped it broken.
+
+`_verify_moves()` iterated `rep.moved` — the keys it had just written — and
+asked whether each read back equal. On a store where the write lands, it always
+does. So a migration handed a `unit` that is on no roster wrote four keys, read
+four keys back, set `verified = True`, and the caller stamped
+`maint_schema_version`. Every symptom of a clean migration, and an install's
+whole maintenance history parked under a name nothing will ever look up. And
+because the marker was stamped, **the next start skipped the migration too** —
+which is precisely the state `store.py`'s header says must never happen: a
+failed migration a later version treats as done.
+
+Reproduced on the reviewer's roster, a house with a garage pack and a shed pack,
+carrying a flat `last_completed` of 2026-08-07:
+
+| | `unit` the migration was handed | key written | pack 0 reads back | due now? |
+|---|---|---|---|---|
+| correct | `self.inverters[0].name` = `garage` | `unit/garage/last_completed` | `2026-08-07` | **False** |
+| wrong | `inverter.UNIT_A` = `inverter-a` | `unit/inverter-a/last_completed` | `None` | **True** |
+
+`build_roster()` names a declared roster's entries **from the config**, and its
+fallback for a nameless entry is a **positional** `inverter-1` — so `inverter-a`
+is not the name of anything on a declared roster, not even by default.
+
+**The remedy, and why this one.** `migrate_maint_keys()` now takes the **roster**
+in place of an `n_units` count (the count was `len(self.inverters)` at the only
+call site, so the roster carries it, and a count passed beside the names is a
+second opinion about how many packs there are). `_verify_moves()` grades the
+namespace it wrote against those names, and a unit that is not among them is
+reported as `misdirected` — a **different** failure from `lost`:
+
+* `lost` is *the store took the write and dropped it* — a storage fault, and it
+  has a harmless-default case (a write of `last_completed: None` loses nothing,
+  because "absent" and "never recalibrated" are the same statement).
+* `misdirected` is *the store did exactly as it was told, by a caller that was
+  wrong*. There is no harmless-default version of it: a write nobody reads is
+  not information, whatever its value. So anything in it fails `ok` on its own.
+
+And **`ok` gates the marker**, so a misdirected migration is not stamped and the
+next start runs it again. The argument for it is asymmetric: a re-run is
+recoverable — rule 1 makes it a no-op where keys did land, and rule 2 left the
+flat keys in the file to recover *from* — while a silent skip is not recoverable
+at all, because the store now claims the migration happened.
+
+### ⚠️ THE PARAGRAPH ABOVE USED TO END HERE, AND AS IT STOOD IT WAS FALSE
+
+**The refusal lasted exactly one start. The second start stamped the marker.**
+That is a correction to what the first fix round wrote down, not a new
+qualification of it, and it is recorded here because a decision document that
+states a guarantee the code does not provide is worse than one that says
+nothing.
+
+`_verify_moves()` graded only `rep.moved`. Measured at Controller level over
+**one persisted store**, roster `["garage", "shed"]`, flat `last_completed`
+2026-08-07, with the call site still handing `inverter-a` on **both** starts:
+
+```
+construction 1: ok=False misdirected=4 moved=4 kept=0   marker=<absent>   (withheld)
+construction 2: ok=True  misdirected=0 moved=0 kept=4   marker=1          (STAMPED)
+history parked under unit/inverter-a/…; marker stamped after ONE restart.
+```
+
+On start 2 the same wrong call site finds the four keys it wrote already
+present, so **rule 1** — never overwrite a namespaced value — routes every one
+of them to `rep.kept`. `rep.moved` is empty, the roster-grading loop iterates
+over nothing, `misdirected` is empty, `ok` is True, the marker is stamped, and
+start 3 takes the `from_version >= MAINT_SCHEMA_VERSION` early return **forever
+— including after the call site is fixed**. Loud-and-recoverable became
+silent-and-permanent after one restart: exactly the state `store.py`'s header
+says must never happen, reached by the fix written to prevent it.
+
+**The remedy is one line**, and it is sound by construction rather than by
+argument: grade `set(rep.moved) | set(rep.kept)`. Every entry in *both* lists is
+a `unit_key(unit, legacy)` built from the **same** `unit` argument, so grading
+`kept` grades the same namespace decision `moved` does. A correct call site's
+`kept` entries are under a roster name and pass; a wrong one cannot launder its
+namespace by having already written it. `migrate_unit_keys()` passes
+`roster=None` and returns before the loop, so it is unaffected.
+
+**Why the check missed it, which is the part worth keeping.** Part 9b's recovery
+check was `migrate_maint_keys(_m5, ROSTER[0], ROSTER)` — the **corrected** name.
+It measures recovery along the one path where the operator has *already fixed
+the bug*. The real failure is a call site that is still wrong on the next start,
+and nothing in the 19 files called `migrate_maint_keys()` twice with the **same
+wrong** `unit`. The check was true, passed, and could not reach the failure its
+sentence named — MUTATION.md's third defect, this time sitting inside the fix
+written to close it. Part **9c** now drives two starts with the same wrong unit
+over one store, asserts the marker is still withheld on the second, and asserts
+the *shape* of the second report (`moved` empty, `kept` full) so it cannot
+quietly revert to 9b's path.
+
+### The marker gate is `rep.ok` and nothing weaker — both guards, each pinned
+
+`if rep.ok:` is `not lost and not misdirected`, and until this round only the
+`misdirected` half was reachable by a check that also had a non-empty `kept`.
+The mutation
+
+```
+if rep.ok:      ->      if rep.ok or rep.kept:
+```
+
+stamps the marker on a migration that **lost a month of history** — defeating
+the older, settled `lost` guard as well as the new one — and **all 19 suites
+were green on it**. Part **9d** is the fixture that separates them: two keys
+already namespaced (`kept` non-empty, so the mutant's `or` fires) and two keys
+whose write the store drops (`lost` non-empty, so `ok` is False), with
+`misdirected` empty throughout so the section cannot pass on the other guard by
+accident. Measured differential on it: shipped → marker absent; mutant → marker
+`1`, both reporting `ok=False lost=2 kept=2`.
+
+**Why the fixtures had to change, and this is the part that got the card
+rejected.** Nothing in the repository could see whether any of the above was
+true, because `inverter.UNIT_A` **is** the string `"inverter-a"` and every
+two-pack fixture in `test_maintenance.py` named its first roster entry
+`inverter-a`. `self.inverters[0].name` and `UNIT_A` were therefore the same
+value in every check that existed, and swapping one for the other at the call
+site left **all 19 suites green**. MUTATION.md's third defect exactly — a check
+whose *data* cannot reach the failure it names — and the fifth consecutive card
+to carry it. Parts 9 and 11 now run on `["garage", "shed"]`, each behind a
+**precondition check** asserting that the first name is not `UNIT_A`, so an edit
+that makes them equal again goes red instead of making the claims vacuous.
+
+### And the flat keys are kept LIVE, which is the half DEV-02 had to add later
+
+Rule 2 originally said the flat keys are left in place and go stale. DEV-02
+amended it, for the sign latch, to *"left in place is not enough"* — and the
+same sentence holds for history. A fossil dated the day the second inverter was
+declared would hand a rolled-back single-FSM add-on a `last_completed` months
+old (it recalibrates a pack that does not need it) or a stale `phase` with a
+`phase_started` old enough to abort on the 12 h timeout the moment it comes up.
+So `UnitHistory` **mirrors the first pack's writes back onto the flat keys** for
+as long as the roster has more than one entry. Only the first pack: the flat
+spelling means `inverters[0]` on every install that has one, and a mirror from
+pack B would overwrite pack A's rollback view with pack B's phase.
+
+## 3. Staggering: the interlock defers, and the stagger is the residue
+
+Two packs on one monthly schedule are due in the same hour every month. The
+options were a stagger **by construction** — offset each pack's first cycle by
+`interval_days / N` — or **deferral**. Deferral, for three reasons:
+
+1. **A constructed stagger cannot remove the need for the interlock.** A forced
+   start, a commissioning cycle, an abort-and-retry, or a pack added mid-month
+   desynchronises any static offset immediately, so the interlock has to be
+   there anyway — and then the stagger is a second mechanism that must also be
+   correct, for a case the first one already handles.
+2. **A deferral is observable and a stagger is not.** The deferred pack says, in
+   the action log and on the status page, that it is waiting, which pack it is
+   waiting for, and that it will run at the next start hour. A silent offset
+   moves an operator's configured start day and explains nothing.
+3. **The stagger comes out of the deferral for free and then persists.** Pack A
+   completes on day 1 and is next due on day 29; pack B, deferred, completes on
+   day 2 and is next due on day 30. One contested month separates them by one
+   cycle length and they stay separated.
+
+**Starvation was the risk and it is closed at two points.** A deferral does
+**not** stamp `last_start_attempt` — that key is the post-abort cooldown, and at
+90 minutes inside a 60-minute start hour, stamping it would lock the deferred
+pack out of exactly the hour it is waiting for. And the winner is the
+longest-waiting pack rather than the first in the roster, so a pack that keeps
+aborting cannot monopolise the slot month after month while its sibling drifts —
+which is this card's own defect wearing a different costume.
+
+**The deferral must not read as a fault.** §12's 12 h and 36 h phase timeouts
+mean *"the battery is not following commands"*. A deferred pack stays `IDLE`
+with `phase_started` untouched, so no timeout is reachable by waiting, and the
+line it logs says WAITING, names the pack it waits for and says it will run
+next — none of which any failure path in `maintenance.py` says.
+
+## 4. Rule 6 (the card calls it rule 8): the first entry's cap is no longer right
+
+`validate.py`'s rule 6 comment said, in as many words, that
+`maintenance_charge_w` was graded against `inverters[0]`'s cap because there was
+exactly one `Maintenance` and it could only ever recalibrate the first pack —
+and that taking `min()` would be an over-refusal, because it would refuse *"a
+5000 W pack being recalibrated beside a 2000 W sibling the cycle never
+touches"*. It closed: **"when `maint_unit` stops being `inverters[0]`, this line
+moves with it."**
+
+It has stopped. Every pack is recalibrated on its own schedule, so **the sibling
+the cycle never touches does not exist** and the over-refusal argument has no
+subject. The shaping claim is filed on whichever pack is charging, so the cap
+that can contradict it is any cap on the roster and the one that contradicts it
+first is the smallest.
+
+**This refuses configurations that used to start**, and that is the point rather
+than a side effect: `max_w` 5000 on pack A beside 2000 on pack B with
+`maintenance_charge_w` 3000 was correct while only pack A was ever recalibrated,
+and is now a cheap-window charge asking the arbiter for `hi = -3000` against a
+supervised `lo = -2000` — `lo > hi`, arbiter clause 3, 0 W, on the pack nobody
+was watching. Refused when it is **written**, which is rule 6's whole job. At a
+roster of one `min()` over one cap is that cap: same comparison, same message,
+same boundary.
+
+## 5. The claim stays per-unit, and nothing here files a site claim
+
+`arbiter.CLAIM_SCOPE` has classified `maintenance` as `SCOPE_UNIT` since it was
+written. Two FSMs mean two claims on two actuators, which the allocator consumes
+as per-unit intervals. [[ARCH-01]] is explicit that this must never become a
+site claim, because per-inverter constraints filed site-wide are the
+cross-inverter `lo > hi` contradiction at `arbiter.py:114`. The only line this
+card changed there is **which** roster entry the claim is filed on: the one whose
+state machine produced the result, rather than `inverters[0]`.
+
+## 6. FOUND AND NOT FIXED: the site loop still contradicts a drain at N>=2
+
+Measured while wiring this card, and **pre-existing on `release/1.0` at
+`2d025ee`** — the same drain filed on `inverters[0]` produces the identical
+refusal with the two unit names swapped.
+
+The maintenance SET **pins** its pack. The loop's target is a **site** claim.
+When the site target is *below* the pin — a 2000 W drain while the house imports
+less than that, which is most of a night — the only split that sums to the site
+target asks the sibling to **charge** while the maintained pack **discharges**,
+and `allocator.plan()` refuses it by name as circulating power. The refusal is
+correct and fails safe: the whole site goes to 0 W with an `ARBITER: REFUSED`
+line in the action log.
+
+**The consequence an operator would see is a misdiagnosis.** The pack is held at
+0 W while its state machine believes it is draining, so after 12 h the DRAIN
+timeout fires and reports *"drain stuck"* — §12's words for a battery that is not
+following commands. The battery is fine; the site refused to allocate.
+
+**Why it is not fixed here.** The fix is to change what the site loop's target
+*means* while a unit is pinned — regulate the meter using only the unpinned
+packs — which is loop and allocator territory and changes the integrator's
+subject. MAINT-05 moves which pack is pinned and nothing else about it. It is
+pinned as a running check (`test_maintenance.py` part 12) rather than left as a
+comment, so it cannot be forgotten or quietly tuned away, and it wants a card of
+its own.
+
+## 7. FOUND AND NOT FIXED: `migrate_unit_keys()` has the same hole, and it is a LATCH
+
+Found while fixing §2's instance. It is the same defect one module over and it
+costs a **disarmed safety guard** rather than a lost history.
+
+`Controller.__init__` calls `migrate_unit_keys(store, UNIT_A)` — a hardcoded
+constant where the maintenance migration now passes the roster's own first name.
+It **has** to run before `build_roster()`, because each `Inverter` restores its
+sign latch in its constructor, so there is no roster to name yet and
+`_verify_moves()` is passed `roster=None`. On a **declared** roster a pre-DEV-02
+`sign_tripped: true` therefore lands under `unit/inverter-a/sign_tripped`, which
+no unit on that site reads.
+
+**And the guard that exists for exactly this is defeated by the key the defect
+wrote.** `_grade_store_migration()` re-arms the latch when the flat key says
+tripped and *nothing namespaced is latched*. The misdirected write **is**
+namespaced and **is** latched, so `namespaced_latched` is `True`, the guard
+returns, and no inverter on the roster comes up tripped — on an install whose
+`/data/state.json` still reads `sign_tripped: true`. That is DEV-02's own
+sentence for the worst case: the state file says protected, the banner is green,
+and the controller resumes commanding into the fault that tripped it.
+
+Measured, on the same `["garage", "shed"]` roster:
+
+```
+pre-DEV-02 flat latch + declared roster ['garage', 'shed'] ->
+    migrated 2 flat key(s): unit/inverter-a/sign_reason, unit/inverter-a/sign_tripped
+latch keys in the store: ['sign_tripped', 'site_sign_tripped', 'unit/inverter-a/sign_tripped']
+roster came up tripped:  [('garage', False), ('shed', False)]
+the same store on the FLAT config -> [('inverter-a', True)]
+```
+
+The last line is the counterfactual: on the flat shape, where `UNIT_A` really is
+the roster's only name, the same store migrates and comes up latched. So this is
+a defect of the **declared roster**, not of the migration as such.
+
+**Why it is not fixed here.** The remedy is to name the migration's target from
+the config before any `Inverter` exists, which means factoring `build_roster()`'s
+naming rule out of `build_roster()` so there is still only **one** spelling of
+it — `inverter.py` and SAFETY-02's latch semantics, with its own mutation battery
+over `test_inverter.py` and `test_watchdog.py`. MAINT-05 moves the maintenance
+history and nothing about the sign latch, and a silent expansion into a safety
+latch is how a rejected PR becomes a broken one.
+
+It is pinned as a running check (`test_maintenance.py` part 13) rather than left
+as a comment. ⚠️ **Its assertions state the WRONG behaviour on purpose** — read
+them as the specification of a defect; the card that fixes it should expect them
+to go red and should replace them.
+
+### And its MARKER GATE had the round-3 hole too — found here, closed here
+
+Separate from the namespace hole above, and **fixed** rather than recorded,
+because closing it cost one test fixture and no source change.
+`migrate_unit_keys()` gates `schema_version` on `rep.ok` exactly as
+`migrate_maint_keys()` gates its own marker, and every check that reached that
+gate ran on a store where **nothing** was namespaced yet — so `rep.kept` was
+empty in all of them and
+
+```
+if rep.ok:      ->      if rep.ok or rep.kept:
+```
+
+**survived all 19 files**. What it does is stamp the schema marker on a
+migration that DROPPED THE LATCH: the guard is off, the state file still reads
+`sign_tripped: true`, and because the store now says the migration finished, the
+next start skips the one thing that would have re-armed it. DEV-02's own worst
+case, filed under *done*.
+
+`test_inverter.py` part 10 now drops the `sign_tripped` write **beside an
+already-namespaced `sign_reason`** — `lost` non-empty *and* `kept` non-empty,
+which is the one shape none of the existing fixtures had — and asserts the
+marker is withheld, with the accepted-baseline counterfactual beside it (habit
+7). The shipped gate was correct throughout; what changed is that a check can
+now see it.

goodwe_controller/DECISIONS/SAFETY-30-per-unit-sign.md

@@ -0,0 +1,474 @@
+# SAFETY-30 — what the sign watchdog compares once there is more than one inverter, and the numbers it compares with
+
+**Status: DECIDED 2026-09-04.** The one comparison becomes **two**, asked of the
+evidence that can answer each: **each unit's standing command against that
+unit's own measured power** (register 0x0518, per-instance since [[DEV-02]]),
+and **the sum of the standing commands against the meter**. At a roster of one
+nothing changes at all — the two questions are the same sentence there, and the
+detector that ships is the only one armed.
+
+This document records the argument for that split, the campaign the four numbers
+were read off, **a number that was carried and a number that refused to be**, and
+the mutation ledger. Filed with SAFETY-30, card `6a9a1828e844f22fabf7fe9c`.
+
+---
+
+## 0. The half that was already built, and the half this card is
+
+[[DEV-02]] shipped the **per-instance** half and nothing here re-tests it: each
+`Inverter` owns its own `SignWatchdog`, the latch is namespaced under
+`unit/<name>/sign_tripped`, the flat pre-DEV-02 pair is maintained as a rollback
+view, and a trip on B no longer latches A. `test_inverter.py` part 3 grades that.
+
+What was left is **the comparison itself**, and it was one line:
+
+```python
+inv.sign_wd.evaluate(stood, self.grid, self.following, now)   # every unit
+```
+
+`stood` is the command **this controller put on THIS actuator**. `self.grid` is
+what the **whole site** is exchanging with the network. At one inverter those are
+two views of one thing — §6.2's identity — which is what makes the check sound
+and what earned its record. At two they are not, and it fails in **both**
+directions:
+
+| | what happens | why no threshold fixes it |
+|---|---|---|
+| **false trip** | A commanded −1500 W while the meter reads +1500 W, because B is discharging 3000 W exactly as told | the meter is answering for a unit A's command says nothing about |
+| **missed fault** | B commanded +1500 W and **measured** at −1500 W, beside A at +1500 W — the meter sits where a healthy site leaves it | the summation **destroys** the evidence |
+
+NOTES.md calls SAFETY-02 the highest-value check in the safety table: 93 %
+agreement while healthy against 4 % during runaway, four observed episodes, zero
+false positives. **That record was earned at N=1 on the site comparison.** The
+first row above spends it; the second row is the guard being absent when it
+matters.
+
+## 1. The two questions, and the two remedies
+
+> **PER UNIT** — this unit's standing command against **its own** measured power.
+> Conclusion: *the actuator I am driving is not doing what I told it.* Evidence
+> about one device, so the remedy is that device: the `failsafe` SET, which
+> `arbiter.CLAIM_SCOPE` has classified `SCOPE_UNIT` since it was written.
+>
+> **SITE** — the **sum** of the standing commands against the meter. Conclusion:
+> *the site as a whole is driving the wrong way* — an inverted gain, an
+> integrator wound past the rail, a strategy on the wrong side of the meter.
+> That is the 2026-08-20 fault day, it is a property of the **connection**, and
+> the remedy is a new SITE SET, `sign_site`, that stops every inverter.
+
+### Why `failsafe` stays per-unit and `sign_site` is new
+
+`CLAIM_SCOPE`'s entry for `failsafe` used to read *"the evidence is site-wide and
+the remedy is not"* — a site measurement attributed to one actuator on the
+strength of the N=1 identity. After this card the first half of that sentence is
+simply true: the evidence **is** per-unit. The site-wide half did not disappear;
+it moved to a claim that is site-scoped for the reason `circulation` is, one
+entry above it: **the offending number is the site target, which no single
+inverter owns.** Parking one unit on it would leave the rest of the roster
+pushing the same way with the guard satisfied.
+
+### At N=1, exactly one detector is armed
+
+`Controller._arm_sign_watchdogs()` is the only place the roster decides anything
+here, and it is a pure function of `len(self.inverters)`:
+
+* **one inverter** — the unit detector is armed against **the meter**, at the
+  shipped `sign_disagree_w` / `sign_disagree_s`. The site detector is
+  **constructed and never evaluated**. Same comparison, same thresholds, same
+  reason string byte for byte, same namespaced keys and flat mirror, same
+  `failsafe` claim, same banner, same single status row. `test_watchdog.py`
+  part 7 asserts each of those, including that the site detector's window was
+  never so much as opened.
+* **two or more** — each unit against its own 0x0518 power under the per-unit
+  pair, and the site detector against the meter under the site pair.
+
+Arming both at N=1 would be two latches and two banners for one fault, and a
+false-positive record measured on **one** comparison spent twice.
+
+## 2. What this is NOT — the two neighbours, named
+
+Two guards that overlap without saying so is how a false-alarm generator gets
+built, so both are written down.
+
+### It is not [[SAFETY-31]]
+
+| | **SAFETY-31 — `circulation.py`** | **SAFETY-30 — the per-unit half** |
+|---|---|---|
+| compares | the units **against each other**: `Σ\|Bᵢ\| − \|ΣBᵢ\|` over measured power | each unit against **its own command** |
+| catches | two inverters pushing **through** each other; nets to zero at the meter | a unit **not doing what it was told**; that unit is faulty or wedged |
+| attributes | nothing — the quantity is defined over the roster and is identically zero for one unit | **the unit**, by name, and parks it |
+| scope | SITE claim, flat latch | UNIT claim, namespaced latch |
+
+They genuinely overlap in **one** scenario — a unit reversing while its sibling
+pulls the other way — and `test_watchdog.py` part 4 asserts exactly that, on the
+same fixture, with both readings printed: circulating power is over its threshold
+**and** SAFETY-30 names B. Neither is the other with a different threshold. Take
+two units faithfully executing an allocation with two signs: circulating power is
+the whole loop and every per-unit sign check reads perfect agreement. Take one
+unit doing the opposite of its share while its sibling covers: this reads the
+whole disagreement and circulating power can be zero.
+
+### It is not the saturation freeze
+
+`compute()`'s anti-windup test is `|prev_w − actual_w| > saturation_w` (500 W). It
+concludes *the inverter cannot follow*, it fires on a merely **saturated** unit as
+readily as on a reversed one, it **stops the integrator winding** — an ordinary
+control action taken every day — and it neither latches nor parks anything.
+
+This asks for **opposition**: a sign, on both sides of a threshold, held
+continuously. A saturated-but-obedient unit cannot produce it. And the two meet
+at a number rather than overlapping in a band nobody drew — see §3.
+
+## 3. The numbers, and where they come from
+
+Read off `admin/goodwe`'s `sim/` two-master rig (TEST-19), used **read-only** and
+run container-free. Real and imported, never transcribed: the control law
+(`control.compute`, `Tuning`), the allocator (`arbiter.allocate_record` →
+`allocator.plan`), the sign predicate (`watchdog.disagrees`), the wire
+(`frame_check.setpoint_atomic`, the firmware's own encoder), the plant
+(`plantsim.Plant`), and the per-unit reading, taken from **holding register
+0x0518** — the register the live `batt_entity` binds.
+
+**1890 closed-loop runs in which NOTHING is faulty** — every unit executes the
+share the allocator gave it — over **6 gains × 5 slews**, the whole range
+`config.yaml` accepts for `gain` `float(0.05,1.0)` and `slew_w` `int(50,5000)`,
+across both families of stimulus that can produce legitimate opposition:
+
+* **1050 load reversals** — 7 sizes (±300 W … ±4000 W) × 5 write staggers, the
+  serialised write path landing unit B's number up to four cycles after A's;
+* **840 flapping loads** — 7 periods (10 s … 600 s) × 4 amplitudes.
+
+Every run is executed **twice**: once at a roster of two and once at a roster of
+one on the same stimulus. **The N=1 arm is the control**, and it is the reason
+this document can separate *"a property of the check that shipped"* from *"a
+property of the second inverter"* instead of asserting one and hoping.
+
+Supervised caps are the shipped `max_w` (2000 W per unit), so the site rail is
+2000 W at N=1 and 4000 W at N=2 — `Controller.site_max_w`, the roster sum.
+
+### Reproducing it
+
+The harness is **not committed here**, for the reason SAFETY-31's was not: it
+imports `plantsim` and `frame_check`, which live in the other repository and are
+not present in CI. What it does is small enough to restate: build two
+`plantsim.Inverter`s from `INVERTER_A` / `INVERTER_B`, close the loop with the
+add-on's own `compute()` and `arbiter.allocate_record()`, put each share on its
+own wire through `Inverter.command()`, advance `Site.settle()` one control cycle,
+and record `(standing command, inv.reg(0x0518))` per unit and
+`(Σ standing, site.grid)` for the site. Then, for each candidate threshold, the
+**longest run of consecutive cycles** on which the shipped `disagrees()` is true —
+not a count of true samples, because the watchdog's window is cleared by any
+disagreeing sample. Point it at a checkout with `GOODWE_ADDON_REPO`.
+⚠️ MUTATION.md's note applies: those harnesses write `__pycache__` into **both**
+trees.
+
+### A. PER-UNIT — a unit's standing command vs its own 0x0518 power, N=2
+
+| threshold | worst continuous, reversals | worst continuous, flapping | runs crossing (of 1890) |
+|---:|---:|---:|---:|
+| 50 W | 25.0 s | 20.0 s | 962 |
+| 100 W | 20.0 s | 15.0 s | 511 |
+| 150 W | 15.0 s | 15.0 s | 379 |
+| 200 W | 10.0 s | 15.0 s | 258 |
+| **250 W** | **10.0 s** | **10.0 s** | **194** |
+| 300 W | 10.0 s | 10.0 s | 124 |
+| 500 W | 5.0 s | 5.0 s | 37 |
+| 700 W | 5.0 s | 5.0 s | 16 |
+| 1000 W | 0.0 s | 0.0 s | 0 |
+
+Largest opposed magnitude anywhere in the 1890 runs: **875 W**.
+
+**The curve is flat, and that is the finding that decided the shape of this
+pair.** Unlike the site comparison — where the meter is driven by the LOAD and a
+flapping load can hold it opposed for minutes — a unit's own power **chases its
+own command** through a bounded ramp and a first-order lag, so the opposition can
+only last as long as the crossing does. Dropping the threshold from 300 W to
+50 W buys 15 s of extra transient and nothing else. **So the work is done by the
+window, and the threshold is put where the control law already draws its line.**
+
+#### `unit_sign_disagree_w: 250` — schema `int(50,5000)`
+
+`saturation_w / 2`. The check gates on **both** sides being over `min_w`, and the
+two are **opposed**, so the smallest command-versus-power gap it can ever see is
+`2 × min_w`. At 250 W that is exactly the 500 W at which `compute()`'s own
+anti-windup test already concludes the inverter is not following. **This watchdog
+therefore cannot fire on a divergence the shipped control law itself calls
+ordinary tracking**, and the two guards meet at a number.
+
+Rejected: a threshold above the observed 875 W maximum, which would gate nothing
+legitimate in — and would also be blind to a unit reversing a 900 W share.
+
+#### `unit_sign_disagree_s: 85` — schema `int(30,1800)`
+
+Three bounds, and the value clears all three.
+
+* **lower, from the rig — 10.0 s.** Two control cycles, the worst continuous
+  legitimate opposition above 250 W across all 1890 runs. It is a reversing site
+  whose second write lands a cycle after the first, both units then slewing
+  through zero behind a 10 s lag. Nothing is wrong.
+* **lower, from the firmware — 30 s (`inverter.LINK_STALE_S`).** A unit not
+  executing its share for less than the firmware's own veto window is a *link*
+  fault the add-on already answers by dropping the unit from the split. Firing
+  inside that window would raise a second, louder alarm about a state already
+  being handled — the SAFETY-15/SAFETY-19 failure exactly.
+* **upper, from SAFETY-02 itself — 180 s, the site hold.** A unit that is not
+  doing what it was told must not be parked *later* than the site would be
+  parked for the same watts, and this conclusion needs nothing to net out: it is
+  complete on the first cycle it appears.
+
+**85 s** is seventeen control cycles: **8.5× the worst legitimate episode**,
+**2.8× the firmware veto**, and **under half** the site hold. Deliberately not
+60, 75, 90, 120 or 180.
+
+### B / C. SITE — Σ standing commands vs the meter, at BOTH rosters
+
+**Load-reversal family** (1050 runs each), worst continuous legitimate opposition:
+
+| threshold | **N=2** | **N=1 (control)** | ratio |
+|---:|---:|---:|---:|
+| 100 W | 335.0 s | 175.0 s | 1.91× |
+| 200 W | 260.0 s | 120.0 s | 2.17× |
+| **300 W** | **215.0 s** | **110.0 s** | **1.95×** |
+| 500 W | 155.0 s | 90.0 s | 1.72× |
+| 700 W | 130.0 s | 70.0 s | 1.86× |
+| 1000 W | 85.0 s | 35.0 s | 2.43× |
+
+**The second inverter roughly DOUBLES the worst legitimate crossing, and at the
+shipped 300 W threshold it takes it from 110 s — comfortably inside the 180 s
+hold — to 215 s, which is outside it.**
+
+The mechanism is not mysterious and is worth stating because it predicts the
+ratio: the transient the 180 s hold exists to reject is the loop crossing zero
+after a load reversal, and the command has to travel the whole **site rail** at
+`slew_w` per cycle. At N≥2 that rail is `Controller.site_max_w`, the roster sum,
+while `slew_w` is unchanged. The worst case is the corner of the config space
+where that takes longest — `gain: 0.05`, `slew_w: 50`, both **schema floors** —
+against a ±2200 W load reversal, which is an oven and a heat pump.
+
+#### `site_sign_disagree_w: 300` — CARRIED, and here is the justification
+
+The argument behind 300 W is about **noise**: *"near zero the sign is noise, and
+near zero is where a healthy loop LIVES"* — the deadband holds the command still
+while the meter wanders either side of `target_grid_w`. `deadband_w` is 15 W and
+there is one P1 whatever is behind it. **Nothing in that argument depends on the
+roster**, and the campaign agrees: the fraction of samples gated in at 300 W is
+within a few percent at both rosters. Carried, with a measurement behind the
+carry rather than an assumption.
+
+#### `site_sign_disagree_s: 325` — NOT carried, and that is the point of the card
+
+180 s is **inside** the 215 s the two-inverter campaign measures, at a
+configuration the schema accepts. Carrying it would have shipped a false trip on
+the check with the best record in the safety table — which is the exact failure
+this card was opened to prevent, arriving from the other direction.
+
+**325 s** is 1.51× the measured worst. It is well inside the **900 s** quarter
+that bills the capacity tariff, and it is inside **every** episode the fault day
+contains: 2026-08-20's four clusters run 16, 30, 77 and 77 minutes, and its
+opposed runs reach 1750 s. What it costs is stated rather than hidden — about two
+and a half extra minutes of wrong-direction **site** power before the stop —
+and it is bought to remove a false trip a legal configuration produces.
+Deliberately not 300 or 360.
+
+**Rejected alternative:** keep 180 s and raise the site threshold to 1000 W,
+where the worst is 85 s. It gives a better margin and a faster stop, and it is
+wrong: it would be raising a **noise** gate to reject a **transient**, which is
+what the window is for, and it would leave the site check blind to a 900 W
+whole-site reversal at a 4 kW site.
+
+### A limitation this campaign found, which SAFETY-30 does not introduce and does not fix
+
+The **flapping-load** family behaves completely differently for the SITE
+comparison, and the control arm is what makes it interpretable:
+
+| threshold | **N=2** | **N=1 (control)** |
+|---:|---:|---:|
+| 300 W | 980.0 s | 1000.0 s |
+| 700 W | 975.0 s | 1000.0 s |
+| 1000 W | 950.0 s | 1000.0 s |
+
+A square wave whose half period is about one control cycle opposes the command on
+**every** cycle: the command chases the previous half-cycle's meter while the
+meter has already reversed. **No window inside the billing quarter survives it and
+no threshold up to 1000 W does either — at EITHER roster.** The N=1 column is
+this document's whole reason for existing: this is a property of comparing a
+lagged command against a meter, and it has nothing to do with the second
+inverter.
+
+It is not acted on here, in order of weight: it is **identical at the roster this
+add-on ships**, so acting on it would be changing a live safety path on evidence
+that has nothing to do with this card; a ~6 kW square wave at 0.1 Hz sustained
+for a quarter of an hour is not a house (the captured healthy day's longest
+opposed run is **30 s**); and the fix is not a number — it wants the detector to
+notice that the disagreement is **alternating**, which is a different check.
+**It is a card, and this section is its evidence.**
+
+### D. The fault arm — because a survey of healthy data is not a derivation
+
+210 runs with inverter B executing the **negation** of its share throughout,
+sibling faithful, meter left to net:
+
+| threshold | longest run the per-unit check detects | runs the SITE check also sees |
+|---:|---:|---:|
+| 250 W | 890.0 s | 210 / 210 |
+| 300 W | 890.0 s | 180 / 210 |
+| 700 W | 885.0 s | 112 / 210 |
+| 1000 W | 885.0 s | 60 / 210 |
+| 1500 W | 880.0 s | 45 / 210 |
+| 2000 W | 0.0 s | 39 / 210 |
+
+At the shipped 250 W the fault is detected continuously for the whole run in
+every one of the 210 — an 85 s hold has ten times the evidence it needs. The
+right-hand column is the honest half: on a *reversing* stimulus the site
+comparison often does see something too, because the reversal is large enough to
+move the meter. **The case it cannot see is the steady one**, and that is the
+fixture `test_watchdog.py` part 4 runs: the shipped comparison is false on
+**0 of 14** cycles while the per-unit check convicts B on every one.
+
+## 4. Inverter B does not exist
+
+Every two-inverter figure in §3 was produced with **inverter A's ramp and lag
+copied onto B** (`plantsim.INVERTER_B`, `measured=False`); its 3000 W rating,
+where this repository's fixtures use one, is a **hypothesis marked as one**.
+[[SPIKE-04]] measures the real ones and is the revisit trigger for all four
+numbers — not a reason to move any of them now.
+
+Which figures depend on it, stated so nobody has to work it out:
+
+* **`unit_sign_disagree_s: 85`** — most. Its lower bound is the 10.0 s crossing,
+  which is a direct function of B's ramp and lag. A slower second unit lengthens
+  it; the 8.5× is what buys headroom for that.
+* **`site_sign_disagree_s: 325`** — yes, but less directly. Its 215 s is
+  dominated by the **rail** and by `slew_w`, which are configuration, not
+  hardware; B's dynamics move it at the margin.
+* **`unit_sign_disagree_w: 250`** — **no.** It is `saturation_w / 2`, read off
+  the add-on's own control law. The rig only confirms that nothing legitimate
+  sits over it for long.
+* **`site_sign_disagree_w: 300`** — **no.** It is the meter's noise floor and the
+  deadband, neither of which is a property of any inverter.
+
+## 5. Mutation ledger
+
+Every run on a **freshly copied tree of its own, no `__pycache__`, mounted
+`:ro`** into the CI image — the structural immunity MUTATION.md records. ⚠️ The
+first battery used ONE shared directory rebuilt between mutations and reported a
+**RED baseline**; a container mounted it mid-copy and read a missing
+`config.yaml` as a failing check. That whole run was discarded: MUTATION.md's
+rule is that a run which could have been racing is not a result, and a battery
+whose baseline is not green establishes nothing. The harness now copies a
+directory per mutation and **refuses to report anything if the baseline is not
+green**.
+
+Byte deltas measured, not estimated; the `+0 B` ones are marked, since a
+size-preserving edit is what TEST-15 turned into a false survivor once already.
+
+| # | file | mutation | Δ | result |
+|---|---|---|---|---|
+| M1 | `main.py` | the per-unit comparison put back to `self.grid` at every roster | −30 B | **RED** — `test_watchdog.py`. This is the card |
+| M2 | `main.py` | the per-unit comparison forced to `inv.batt` at N=1 too | −31 B | **RED** — AC 8 |
+| M3 | `main.py` | the site half never evaluated | −6 B | **RED** — AC 3 |
+| **M4** | `main.py` | **the site half armed at N=1 as well** | **+0 B** | **SURVIVOR on the first battery — GAP CLOSED**, then RED. See below |
+| M5 | `main.py` | the `sign_site` claim never filed | −6 B | **RED** — `test_watchdog.py`. ⚠️ `test_peakguard.py` stayed **GREEN**, correctly: 10f's AST scan pins which sources CAN be filed and the append is still written; whether it is REACHED is a dynamic question no fixture there gets to |
+| M6 | `main.py` | the site latch not persisted to the store | −58 B | **RED** — the flat-keys and restart checks |
+| M7 | `main.py` | the operator clear leaves the site latch set | −42 B | **RED** — the clear check |
+| M8 | `watchdog.py` | `arm()` clears `disagree_since` | +35 B | **RED** — "re-armed on every one of 38 cycles still trips" |
+| M9 | `watchdog.py` | the reason hard-codes "a meter reading" again | +1 B | **RED** — `test_watchdog.py`. `test_inverter.py` stayed **GREEN**; its part-3 row reads the noun off `sign_wd.against` rather than out of the reason string, so the two checks bite in different places |
+| M10 | `watchdog.py` | `unit_sign_disagree_w` 250 → 300 | **+0 B** | **RED** — the `saturation_w / 2` pin and the CFG-03 pin |
+| M11 | `watchdog.py` | `site_sign_disagree_s` 325 → 180, i.e. **the number carried** | **+0 B** | **RED** — the measured-215 s check and the CFG-03 pin |
+| M12 | `arbiter.py` | `sign_site` scoped `SCOPE_UNIT` | **+0 B** | **RED** — `test_arbiter.py` **and** `test_peakguard.py` |
+| M13 | `validate.py` | the `unit_sign_disagree_w` rule removed | −126 B | **RED** — `test_validate.py` |
+| **M14** | `main.py` | **`arm()` handed `self.sign_cfg` at EVERY roster** — the WIRING, not the values | −31 B | **SURVIVOR at review — GAP CLOSED**, then RED. See below |
+| M15 | `main.py` | per-unit `arm()` given the right `min_w` and the SHIPPED hold — a HOLD-ONLY mis-wire | +78 B | **RED** — the same wiring pin |
+| M16 | `main.py` | per-unit `arm()` given the right `min_w` and the SITE hold — hold-only again, and behaviour-preserving | +83 B | **RED** — the wiring pin, and the ONLY red in `test_watchdog.py` |
+
+### M4 is recorded as a survivor because it was one
+
+The check it should have gone red on read:
+
+```python
+not _c1s.site_sign_wd.tripped and _c1s.site_sign_wd.disagree_since is None
+```
+
+Arming the site detector at a roster of one left that **green**, and the reason
+is worth reading because the check looked airtight. In the N=1 fixture the unit
+detector trips first at its 1 s hold and parks the inverter; the site's standing
+command is then **0 W**, so the site predicate is false, its window is never
+opened, and `disagree_since is None` is true *whether or not the detector was
+evaluated at all*. **The assertion ran, the sentence was true, and the data
+could not reach the failure it named** — MUTATION.md's third failure mode,
+inside the check written to prevent it.
+
+Replaced with a **call counter installed on the object** before the run: twelve
+cycles, zero calls, printed. A counter can only be zero for one reason. The
+mutation was re-run against it and is red; the run that found it was discarded
+along with its tree, so this is a result and not an artefact.
+
+### M14: the VALUES were pinned and the WIRING was not
+
+M10 and M11 mutate the numbers in `watchdog.SIGN_OPTION_DEFAULTS` and both are
+red through the config pins. Neither of them touches the line that DELIVERS
+those numbers to a detector:
+
+```python
+inv.sign_wd.arm(self.sign_cfg if one else self.unit_sign_cfg,
+                AGAINST_METER if one else AGAINST_OWN_POWER)
+```
+
+Drop the conditional — `arm(self.sign_cfg, ...)` — and every unit is armed with
+SAFETY-02's meter-facing pair instead of the re-derived per-unit one. At review
+that left **all 19 files green**. `grep -rn "sign_wd.cfg" test_*.py` found
+exactly one threshold assertion, and it is §7's N=1 arm — `(300.0, 1.0)` —
+which the mutation PRESERVES. Nothing anywhere read `inv.sign_wd.cfg` at a
+roster of two.
+
+The behavioural checks in part 3 cannot close it and it is worth saying why: B
+is reversed at ~2000 W and every hold in that fixture is short, so B trips under
+either pair. **A fault big enough and long enough to trip under both thresholds
+is evidence about neither.**
+
+What the survivor costs in the field is the whole product of §3: at
+`config.yaml`'s own defaults the per-unit check would run at `sign_disagree_s`
+**180 s instead of 85 s** — 2.1× slower to park a reversed inverter — and its
+gate would move from 250 W to 300 W, so a unit reversing a share between the two
+becomes invisible, while `SIGN_OPTION_DEFAULTS`, `config.yaml` and §3 all go on
+describing a pair the running code does not use. With CI green.
+
+**Closed by a wiring pin in `test_watchdog.py` part 3**, on the real
+two-inverter Controller: every `inv.sign_wd.cfg` equals the per-unit pair,
+`site_sign_wd.cfg` equals the site pair, and each detector's `against` is its
+own instrument — all graded **against literals**, since `== self.unit_sign_cfg`
+cannot tell *the per-unit pair* from *whatever this Controller was handed*.
+
+⚠️ **And the fixture had to be changed for that pin to mean anything.**
+`_sign_site` set `unit_sign_disagree_s: 1` **and** `site_sign_disagree_s: 1`
+**and** `sign_disagree_s: 1`, with two of the three magnitude gates at 300 W. A
+bare equality there discriminates on `min_w` alone: a mis-wiring that delivered
+the right magnitude with the wrong hold would compare EQUAL to the pair it was
+supposed to have. That is this document's own §5 lesson and MUTATION.md's third
+failure mode — a check comparing two things that happen to be equal cannot see
+the difference it names — reappearing inside the fix for it. The three pairs are
+now pairwise distinct **in both coordinates** (250 W / 1 s, 320 W / 2 s,
+300 W / 3 s), a PRECONDITION check asserts that distinctness so a later edit
+collapsing them goes red rather than quietly vacuous, and both pairs are
+**printed**. M15 and M16 exist to show the hold half actually bites: right
+magnitude, wrong hold, red both times — and M16, whose wrong hold still trips
+inside the drive, is red on the wiring pin and on **nothing else** in that file.
+
+## 6. What this card does NOT close
+
+* **The flapping-load blind spot in the SITE comparison.** §3, at length. Present
+  identically at the shipped roster of one; wants an alternation test, not a
+  number; a card of its own.
+* **A unit misbehaving while it is NOT loop-driven.** Gate 1 (`following`) is
+  kept for the per-unit comparison, deliberately: the whole no-false-positive
+  argument is built on it, and the saturation freeze already grades that window.
+  So between the start of a maintenance cycle and its end, a unit reversing
+  against its own command is the freeze's to notice. Widening it is a card with a
+  rig campaign behind it, not an edit.
+* **A sign inversion downstream of this add-on** — in the firmware, on the wire,
+  or in a future driver. Unchanged: our view stays self-consistent and both
+  comparisons stay silent. That is the witness's problem and it always was.
+* **A false-alarm interaction with SAFETY-31 that nobody has to think about.**
+  They overlap in one scenario and both fire in it. That is correct — they
+  conclude different things — but an operator will see two alarms for one event.
+  Whether the banner should say so in one sentence is a UI card.

goodwe_controller/DECISIONS/SAFETY-32-per-pack-bms-floor.md

@@ -0,0 +1,816 @@
+# SAFETY-32 — `bms_soc_floor` per pack, and the two judgement calls it left open
+
+**Status:** decided and implemented on branch `SAFETY-32`.
+**Card:** SAFETY-32 — *"One `bms_soc_floor` for two packs grades the shallower
+pack against the deeper one's cut-out"*, scoped by the card's own PM PROOFREAD
+to its **third** bullet only. Bullets 1 (aggregate anti-windup) and 2 (per-instance
+`achievable_w`) were verified as already shipped by DEV-02 and CFG-04 and were
+**not** touched — `main.py`'s `actual_w=self.site_batt_w` comment is the record
+of why the first of those must stay site-level, and it still reads as written.
+
+---
+
+## 1. The defect, stated in the terms the code uses
+
+`bms_soc_floor` **describes a battery**: it is the percent that pack's own BMS
+cuts out at, register `0x0560` on the reference GW5000S-BP, read as 10. It was a
+flat option because there was one pack.
+
+`peak_soc_floor` **describes the site's policy**: the depth a money-saving
+discharge may never reach. Rule 8 refuses `peak_soc_floor <= bms_soc_floor`,
+because such a floor can never bind — the BMS wins and the status page reports a
+backstop that does not exist.
+
+At two packs of different models the two cut-outs differ, and one number for both
+is wrong **in the direction that empties a pack**: with cut-outs of 10 % and 30 %
+and `peak_soc_floor: 20`, the add-on **started**, and the guard would command a
+forced discharge that pack B's *BMS* — not the operator's floor — is the thing
+that stops. That is rule 8's own failure mode, reached through rule 8.
+
+---
+
+## 2. DECISION 1 — required per entry, not defaulted from `PEAK_OPTION_DEFAULTS`
+
+The card left this open and asked for the reason to be recorded.
+
+**Decided: REQUIRED of every roster entry**, in `config.yaml`'s schema
+(`bms_soc_floor: int(0,50)`, no trailing `?`) and in `validate.py` rule 8, which
+refuses a missing one **by name**.
+
+The reason is that a default here is not a convenience, it is **this card's own
+defect wearing a different hat**. A default is one number applied to every pack,
+which is exactly what the flat key was. An add-on that answers "10 %" for a pack
+nobody has told it about has invented a property of somebody's battery, and the
+failure mode of the invention is the one this card exists to remove — the
+shallower pack protected by a floor that belongs to the deeper one. CFG-03's
+rule ("defaults live in exactly one place") points the same way: the one mirror
+`PEAK_OPTION_DEFAULTS["bms_soc_floor"]` is the default for the **flat** key, i.e.
+for the single-inverter shape, and reusing it per entry would make one dict the
+default in two shapes with two meanings.
+
+It is also the treatment `max_w` already gets, for the same reason and by the
+same argument (`inverter.py`'s `build_roster()` docstring): the per-unit numbers
+that describe hardware are required, and the per-unit *sensors* are optional
+because an absent sensor is a real state.
+
+**The cost is stated rather than hidden.** This is a schema change: a roster
+config written before this branch is refused by the Supervisor until each entry
+gains the key, and every roster fixture in the suite needed it. Nothing in
+production carries a roster today — `inverters:` ships `[]` and the second
+inverter is ~8 days out — so the migration is paid by fixtures and by whoever
+writes the first real roster, at a startup refusal that names the entry.
+
+**What is NOT inherited.** An entry that omits the key does not borrow the flat
+one. `Inverter.bms_soc_floor` is then `None`, `peakguard.pack_floor_pct()`
+answers `None`, and `discharge_permitted()` already treats an unknown floor as
+**no permission** — so even in the window before rule 8 runs (e.g. `sim/inject.py`,
+which does not run the startup gate) the absence costs a shave and never a pack.
+`0.0` was rejected explicitly: it is the most permissive value the key has.
+
+## 3. DECISION 2 — the site verdict is the MAXIMUM of the packs' own floors
+
+`Controller.pack_floor_soc` — the **minimum** measured SoC, the emptiest pack —
+**is unchanged**, exactly as the card instructed. What changed is the floor it is
+graded against. `peakguard.binding_soc_floor()` reduces the roster's per-pack
+floors, each of them `max(peak_soc_floor, that pack's own cut-out)`, with a
+`max()`. The two reductions pair: a minimum SoC against a maximum floor is the
+conservative combination, and both move the verdict towards refusing a shave and
+away from spending a pack. `min(soc) <= max(floor)` is implied by "some pack is
+at or below its own floor", so the site verdict is never more permissive than
+grading each pack separately would be. The converse is not true and is not
+claimed — a site can be refused while every pack individually has headroom, which
+costs a shave.
+
+### 3a. ⚠️ FIX ROUND — the reduction was written and argued, and NOT GRADED
+
+PR #48 was rejected because the argument in §4 below justified *writing* the
+reduction and was then not carried into a check. Neither of the two mutations
+below appears anywhere in §6's twelve-row battery: it graded *"the reduction
+dropped"* and *"the bare option handed over"*, and never asked what happens when
+the reduction is the **wrong one**, or when the roster handed to it is
+**sliced**. Both were applied to `app/` by the reviewer and **all 19 files
+stayed green**:
+
+| mutation | Δ | why it survived |
+|---|---|---|
+| `binding_soc_floor()`: `max(floors)` → `min(floors)` | **0 bytes** | every roster in the suite was rule-8-valid or one pack long, so `floors` held **identical numbers** and `max`/`min`/first/last were four spellings of one answer |
+| the `decide()` call site: `self.packs` → `self.packs[:1]` | **+4 bytes** | part 12g asserted `"self.packs" in <the unparsed expression>` — and `self.packs[:1]` **contains that substring** |
+
+The zero-byte one was proved to have actually run rather than assumed:
+`dis.get_instructions` showed the `LOAD_GLOBAL`s as `[float, pack_floor_pct,
+any, min]`, and `binding_soc_floor([Pack("a",50,10), Pack("b",50,35)], 20.0)`
+returned `20.0` where the shipped code returns `35.0`.
+
+**The common cause is this repository's recurring defect** — a check whose DATA
+cannot reach the failure it names. The one check that used a cut-out **above**
+the EMS floor (12c's `_n1_deep`) ran on a roster of **ONE**, where every possible
+reduction agrees. So no check in the suite ever evaluated
+`binding_soc_floor()` on two packs whose floors **differ**.
+
+**And the rule-8 degeneracy is not a defence, by §4's own words.** If
+"unreachable is a claim about today's rule ordering, not a property of this
+function, and `validate_startup_config()` is also `sim/inject.py`'s entry point"
+is enough to *write* the reduction, it is enough to *grade* it. Through
+`sim/inject.py`, which does not run rule 8 at all, `min(floors)` lets the guard
+command a forced discharge to 20 % on a pack whose BMS cuts out at 35 %.
+
+**What was added — no shipped behaviour changed, only what observes it:**
+
+* **`test_peakguard.py` part 12c2** evaluates `binding_soc_floor()` on a roster
+  whose per-pack floors **all differ**. Four **simulated** cut-outs —
+  `30, 25, 40, 35` %, every one strictly above `peak_soc_floor: 20`, so
+  `pack_floor_pct()` selects the cut-out for each and `floors` *is* that tuple in
+  that order. The four candidate reductions are then four different numbers:
+
+      max   = 40   <- the shipped answer
+      min   = 25
+      first = 30   <- "grade the site against pack A", inside the function
+      last  = 35
+      mean  = 32.5
+
+  The extremes sit at **interior** roster positions (index 2 is the max, index 1
+  the min). A two-pack roster structurally cannot separate these — there `max`
+  and `last` coincide and so do `min` and `first` — and neither can three. Four
+  is the smallest roster that separates all four. The precondition is its own
+  check (habit 1), one pack is then raised past the old maximum (habit 7: the
+  **last** pack, which was neither extreme, so the line can only move if the
+  reduction read it), and `peak_soc_floor` is lifted above every cut-out to grade
+  the other side of the conjunction.
+
+* **Part 12g is behaviour now, not source text** (MUTATION.md habit 4). Real
+  `Controller.cycle()`s on a two-pack roster, both packs held at the same SoC and
+  everything else identical, with the packs' own cut-outs as the only thing
+  varied:
+
+      cut-outs A 10 / B 18 % -> decide() got soc_floor=20 %, floor_limited=False
+      cut-outs A 10 / B 35 % -> decide() got soc_floor=35 %, floor_limited=True
+
+  The observable is `PeakDecision.soc_floor` — the argument, echoed back by the
+  shipped `decide()` — so the whole path from `Controller.inverters` to the
+  guard's argument is under test, and the site's **verdict flips** as well as the
+  number. `self.packs[:1]` answers 20 % both times. **Which** option key the floor
+  comes from stays pinned on the source in part 4, because that claim genuinely
+  is about the code's shape; that the option still governs is now graded
+  behaviourally too.
+
+* ⚠️ **A THIRD SURVIVOR, FOUND WHILE FIXING THE FIRST TWO, AND RECORDED RATHER
+  THAN QUIETLY REPAIRED.** The first cut of part 12g varied only **pack B's**
+  cut-out — so `self.packs[-1:]`, grade the site against the **last** pack,
+  agreed with the full roster on both runs and stayed green across all 19 files
+  (+5 bytes). A two-run fixture that only ever moves the second pack pins `[:1]`
+  and not `[-1:]`; it is the same defect one step along, and it was in the fix
+  for this defect. Part 12g now runs the **same pair of cut-outs in both
+  orders**:
+
+      cut-outs A 10 / B 18 % -> soc_floor=20 %, floor_limited=False
+      cut-outs A 10 / B 35 % -> soc_floor=35 %, floor_limited=True
+      ...the SAME PAIR SWAPPED, A 35 / B 18 % -> soc_floor=35 %, floor_limited=True
+
+  `[:1]` answers 20 % on the second run, `[-1:]` answers 20 % on the third, and
+  the roster answers 35 % on both. No single-position slice passes all three.
+  This is the mirror `12h` already ran on the SoC axis, which is where the shape
+  was borrowed from.
+
+* ⚠️ **AND A FOURTH, from the same battery: the EMPTY-roster branch.** Rewriting
+  `binding_soc_floor([])`'s `float(soc_floor)` to `return None` (−43 bytes) was
+  green across all 19 files. It is defensive code — `Controller.inverters` is
+  never empty on a started controller, and `inputs_ok` refuses the cycle before
+  the guard is reached if it were — so nothing in the suite went near it. It is
+  **pinned rather than argued**, in both directions (the option when there is
+  one, `None` when there is not), because a branch with no check is a branch
+  that can be rewritten under a green suite. The check also records why it
+  cannot permit anything either way: `pack_floor_soc` is `None` at an empty
+  roster, and `discharge_permitted()` refuses on a `None` SoC whatever the floor
+  is.
+
+* **`wired_over_limit()` gained a `prepare(ctl)` hook**, inside its retry loop so
+  every attempt grades the same site, which is the only window in which a second
+  inverter can join the roster before the telegrams are fed.
+
+* **In-process mutation rows 22, 22b and 23** record all four survivors, so a
+  later edit that re-degenerates any of these fixtures goes red on a row rather
+  than on nothing.
+
+### 3b. ⚠️ SECOND FIX ROUND — the fix carried the defect it was written to close
+
+Round 2 closed both named survivors, and both of round 3's blockers are the
+**same defect sitting inside that fix**: a check whose TEXT claims more than its
+FIXTURE can observe. Applied to `app/` on `8501e05`, **all 19 files stayed
+green** for both:
+
+| mutation | Δ | the sentence that was too broad | why the fixture could not see it |
+|---|---|---|---|
+| `binding_soc_floor()`: `max(floors)` → `max(floors[-2:])` | **+5 bytes** | *"So the reduction reads **every entry**"* | 12c2's only counterfactual raised the **last** pack (`_deeper[-1]`), and **every window that ends at the roster's end follows the last pack**. On `(30,25,40,35)` both answer 40; on `_deeper` both answer 45. |
+| the `decide()` call site: `self.packs` → `self.packs[:2]` | **+4 bytes** | *"what is reduced is the **ROSTER** and not a POSITION in it"* | all three `_floor_reaching_decide` runs drove a **two-pack** roster, where `self.packs[:2]` **is** `self.packs` — by exactly the argument 12c2 itself makes about two packs being unable to separate max/min/first/last. |
+
+Round 2's *report* stated the narrow, true claim ("no **single-position** slice
+passes all three runs"). The *check in the file* stated the broad one. That gap
+is the whole of round 3.
+
+**THE GENERAL LAW, stated once so the next card can use it:**
+
+> **A fixture of N packs cannot separate two reductions that agree on N packs.**
+> Two cannot separate max from min from first from last. Three cannot separate
+> max from max-of-the-last-two. A single counterfactual that varies one position
+> cannot separate the reduction from any window containing that position.
+
+Its corollary is the reason this section is long: **the regress does not
+terminate in a fixture.** Closing `[:1]` took two packs; `[:2]` took three;
+`[:3]` takes four — and a four-pack run hands `[:4]` the same free pass. Round 3
+therefore does two different things rather than adding one more pack.
+
+**What was added — no shipped behaviour changed, only what observes it:**
+
+* **Part 12c2 gained a ROSTER SET and a computed FAMILY TABLE.** The same four
+  simulated cut-outs with the **deepest moved to each of the four positions**,
+  and fourteen enumerated rewrites of `max(floors)` each checked against every
+  roster. The table is **computed and printed**, and a member with an empty kill
+  list takes the check **red naming it** — so a family member can no longer be
+  added without a fixture that reaches it. `max(floors[-2:])` is killed only by
+  the rosters whose deepest sits at index 0 or 1, which is precisely the data the
+  single last-pack counterfactual could not supply.
+
+* **Part 12c2 gained a LENGTH SWEEP.** `binding_soc_floor()` is a pure function
+  over a list, which is the one place a sweep is cheap: **78 simulated rosters,
+  lengths 1 through 12, the deepest cut-out at every position of each.** That
+  closes *every* head window `max(floors[:w])` and *every* tail window
+  `max(floors[-w:])` of width **1 to 11** in one property rather than one
+  mutation at a time — and the check asserts that width **12 is NOT reached**, so
+  the limit is a red line rather than a silence.
+
+* **Part 12g drives THREE-pack rosters**, with the deep cut-out at index 0, 1
+  **and** 2. Every positional slice of a three-pack roster is a proper subset of
+  `{0,1,2}` and so drops a position; the run whose deep pack sits there answers
+  the option, 20 %, where the roster answers 35 %. `self.packs[:2]` goes red on
+  the index-2 run. `_floor_reaching_decide()` now takes `*floors`, and the extra
+  units join through the same `second_inverter()` hook the two-pack runs use.
+
+* **Part 12g gained a POSITIONAL-SLICE TABLE**, computed against the **real**
+  `Controller.packs` each of the six cycles actually ran and the **real**
+  `PeakDecision.soc_floor` it produced — so it measures the shipped call site's
+  own answer, not a re-derivation, and a survivor takes the check red.
+
+* ⚠️ **AND A FIFTH AND SIXTH SURVIVOR, FOUND BY ROUND 3'S OWN BATTERY AND
+  CLOSED STRUCTURALLY RATHER THAN BY ONE MORE PACK.** `self.packs[:3]` (+4) and
+  `self.packs[-3:]` (+5) survive **all six** behavioural runs, because the
+  longest roster those runs drive is three packs long. Part 12g therefore also
+  pins the roster **argument's identity** on the AST: it must unparse to exactly
+  `self.packs`, which is false of every slice at every length. ⚠️ This is **not**
+  the substring test round 2 removed — `"self.packs" in expr` is *true* of every
+  slice; `ast.unparse(arg) == "self.packs"` is *false* of all of them. The two
+  halves are separated and each says what it reaches: the **runs** grade the
+  VALUE and reach every slice up to N=3; the **pin** grades the ARGUMENT'S
+  IDENTITY at any N and is blind to two things the runs are not — a rewrite of
+  what `Controller.packs` *returns* (the runs' `len(ctl.packs) == 3` catches
+  that) and the reduction *inside* `binding_soc_floor()` (12c2's).
+
+* ⚠️ **AND ONE ROW IS RECORDED AS NOT CAUGHT.** A reduction over a window
+  **twelve packs wide** is indistinguishable from `max(floors)` on every roster
+  in the file, because the sweep stops at twelve. Unlike the call site there is
+  no identity claim to pin — the claim is semantic, "the maximum", and a shape
+  pin would refuse honest refactors of the expression. It is reported **with its
+  size** (it needs a 13-pack roster, on a site whose roster is ONE and whose
+  second unit is ~8 days out) rather than hidden.
+
+* **Two rewrites are EQUIVALENT, not uncaught.** `max(set(floors))` and
+  `list(reversed(self.packs))` cannot change any answer — `max` is order- and
+  duplicate-invariant. The second is nonetheless red on the structural pin.
+
+* **In-process mutation rows 24 – 27** record all four of round 3's findings,
+  including the not-caught one, so a later edit that re-degenerates any of these
+  fixtures goes red on a row rather than on nothing.
+
+## 4. ⚠️ THE HONEST LIMITATION, and it is the reason this record exists
+
+**Under any configuration rule 8 accepts, `binding_soc_floor()` answers
+`peak_soc_floor` for every pack.** Rule 8 refuses `peak_soc_floor <= bms_soc_floor`
+on *every* unit, so `max(peak_soc_floor, bms_i) == peak_soc_floor` everywhere the
+add-on will actually start. The per-pack **floor** is therefore non-degenerate
+only against a config rule 8 refuses.
+
+This is not a defect and it is not hidden. It is the same argument
+`validate.unit_caps_indexed()` makes about itself, in that function's own words:
+
+> *"Rule 9(d) already refuses such a config so the two lists cannot desynchronise
+> in shipped use — but 'unreachable' is a claim about today's rule ordering, not
+> a property of this function, and `validate_startup_config()` is also
+> `sim/inject.py`'s entry point."*
+
+CFG-04's index desynchronisation is what happened the last time that reasoning
+was left implicit. So the reduction is written, and it is graded on both sides:
+`test_peakguard.py` part 12c sweeps **1775** rule-8-valid `(peak_soc_floor,
+bms_soc_floor)` pairs and asserts the answer is the option in every one of them
+(that is AC 5, measured over the whole option space rather than at the shipped
+point), and then asserts that on the pair rule 8 *refuses* the answer is the
+**cut-out** — because without that second line the sweep would be satisfied by
+`soc_floor` typed straight through, and the mutation that deletes the `max()`
+could never bite. It does bite: M2 on the PR.
+
+**What IS live at a rule-8-valid roster** is the rest of the card: rule 8 now
+refuses the two-model config that used to start; the permission is answered per
+pack, so the alarm names the pack that stopped the site rather than "the
+battery"; and an entry with no stated cut-out is withheld instead of defaulted.
+
+## 4b. ⚠️ ROUND 4 — the sibling argument, and SPIKE-04 making this the live axis
+
+Rounds 2 and 3 hardened the **floor** axis: the `max` inside
+`binding_soc_floor()`, the roster handed to it, `Controller.packs`, and
+`pack_floor_pct()`'s conjunction. Round 4's blocker is the same family at the
+**sibling argument of the same call**:
+
+```python
+self.peak_guard = peakguard.decide(
+    ...,
+    soc=self.pack_floor_soc,                      # MINIMUM  <- round 4
+    soc_floor=peakguard.binding_soc_floor(self.packs, ...),   # MAXIMUM  <- rounds 2-3
+```
+
+`peakguard.py`'s own docstring names them as a pair — *"MAXIMUM here pairs with
+the MINIMUM in `pack_floor_soc`"* — and a conjunction is exactly as strong as
+its weaker half. On `52380c0`, with all 19 files green and verified again after
+the rebase onto `3934ad7`:
+
+| mutation in `Controller.pack_floor_soc` | bytes | before round 4 | now |
+|---|---|---|---|
+| `min(float(v) for v in vals[:2])`  | +4 | **GREEN** | RED |
+| `min(float(v) for v in vals[-2:])` | +5 | **GREEN** | RED |
+| `min(float(v) for v in vals[:3])`  | +4 | **GREEN** | RED |
+| `vals[:1]`, `vals[-1:]`, `vals[1:]`, `vals[:-1]`, `max(vals)`, the mean | | RED | RED |
+
+The suite's separating power on the SoC axis stopped at exactly **N = 2** —
+precisely the state the floor axis was in before round 3, and which round 3
+itself named as its blocker 2. A fixture of N packs cannot separate reductions
+that agree on N packs.
+
+**It changes a real answer.** A three-pack roster at 50 / 60 / 11 %, emptiest
+pack **third**: the shipped property answers 11 %, `vals[:2]` answers 50 %. The
+site is then graded for discharge *permission* against a pack 39 points fuller
+than the emptiest, and spends a pack its own BMS stops — through
+`sim/inject.py`, which runs no validator rule at all. The verdict flips with it:
+11 % is below the 20 % EMS floor, so the shipped site is floor-limited and sizes
+no forced discharge, where the mutant is permitted and sizes a positive one.
+
+### ⚠️ SPIKE-04 makes the differing-floor discipline the entire exposure
+
+SPIKE-04 reads pack B as **4 × Dyness DL5.0C, DoD 90 → floor 10 %** — **FROM THE
+DATASHEET AND THE SERIAL. Inverter B does not physically exist (~8 days out) and
+nothing about it has been measured.** Pack A is DoD 90 → 10 % as well, and *that
+one is measured*, off register `0x0560` on the reference GW5000S-BP.
+
+So **production is degenerate on the floor axis**: both packs floor at the same
+value, `max(10, 10) == 10`, and a differing floor exists **only in the
+fixtures**. That is not an argument for simplifying the fixtures to matching
+values — it is the argument for keeping them. `binding_soc_floor()`'s reduction
+is observable *nowhere else*, and the moment a third pack, a replacement pack or
+a different chemistry arrives it is observable in the field. **Anyone who
+"simplifies" `A_PACK_FLOOR != B_PACK_FLOOR_SIM` to one number deletes the only
+data this repository has that can see the reduction at all**; part 12c2's
+precondition check goes red if they do, and this paragraph is why.
+
+The **SoC axis has no such degeneracy**. Two batteries at the same state of
+charge is a coincidence, not a configuration, so `pack_floor_soc` is the half of
+the pair that will be non-degenerate in the field from the day inverter B is
+wired up.
+
+### The treatment, transferred from round 3
+
+* **the roster SET** — six real `Controller.cycle()` runs, the emptiest pack at
+  every position of a three-pack roster and both positions of a two-pack one,
+  the observable read off `PeakDecision.soc` (the argument, echoed) rather than
+  off a spy or a grep. Cut-outs held flat across all of them, so the floor axis
+  cannot be what moved the answers.
+* **the positional-slice table and the non-slice table** — computed against the
+  real `Controller.inverters` each cycle carried and the real decision it
+  produced, printed, with a member that nothing kills taking the check red.
+* **the length sweep** — the real property on real rosters of 1 to 12 inverters,
+  the empty pack at every position. Closes head and tail windows of width 1-11
+  at once. Width 12 and wider is **honestly not reached**, and asserted to be
+  unreached so a later sweep that lengthens the rosters must rewrite the
+  sentence.
+* **the exact structural pin** — three of them, because there are three
+  reductions over roster length in this one property: `vals` is built by one
+  comprehension over the bare `self.inverters`; `any(... for v in vals)` reduces
+  the whole list; `min(... for v in vals)` reduces the whole list. Every slice
+  of either is red at **every** roster length, including the `[:3]` and `[:12]`
+  the runs and the sweep cannot see.
+
+Unlike the floor axis, **there is no uncaught row here**: the claim at
+`pack_floor_soc` is an *identity* ("this list"), so a shape pin is the right
+instrument, where `binding_soc_floor()`'s claim is *semantic* ("the maximum")
+and a shape pin there would refuse honest refactors. That asymmetry is why one
+carries an open row and the other does not.
+
+## 4c. ⚠️ THE ENUMERATION — what OTHER reductions over roster length are graded
+
+This card has now been rejected three times for the same shape one step over, so
+round 4 stopped closing members and enumerated the **family**. Every reduction
+over `self.inverters` / `self.packs` in `Controller` was found from the AST and
+mutated to a slice, one at a time, against the full 19-file suite.
+
+| # | site | reduction | consumer | mutation | result |
+|---|---|---|---|---|---|
+| 1 | `pack_floor_soc` | `min` over SoC | `decide(soc=)`, `arbitrage.decide(soc=)` | `vals[:2]` +4 | **was GREEN → now RED** (12i) |
+| 2 | `pack_floor_soc` | `any(v is None)` | same | `vals[:1]` +4 | RED (raises in 12d; pinned in 12i) |
+| 3 | `pack_floor_soc` | the `vals` comprehension | same | `self.inverters[:2]` +4 | RED (12i sweep + pin) |
+| 4 | `binding_soc_floor` | `max` over floors | `decide(soc_floor=)` | `min`, windows | RED (12c2, round 2-3) |
+| 5 | call site | the roster argument | same | `self.packs[:1..3]` | RED (12g, round 3) |
+| 6 | `Controller.packs` | the comprehension | `binding_soc_floor`, `pack_permissions` | `self.inverters[:2]` +4 | RED (12g three-pack runs) |
+| 7 | `site_batt_w` | `sum` + `any(None)` | `decide(batt_w=)` — sizes the floor | `self.inverters[:1]` +4 | RED (test_inverter.py) |
+| 8 | `site_rating_w` | `sum` of per-unit conjunction | `decide(achievable_w=)`, freeze rail | `[:1]` +4 | RED (test_inverter.py) |
+| 9 | `site_max_w` | `sum` of `max_w` | the site loop's supervised bound | `[:1]` +4 | RED (test_inverter.py) |
+| 10 | `site_dev_bound` | `sum` + `any(None)` | `looks_frozen`, `rail_unreachable` | `[:1]` +4 | RED (test_inverter.py) |
+| 11 | `site_circulating_w` | `circulating_w` over measured power | SAFETY-31 trip | `[:1]` +4 | RED (test_watchdog.py) |
+| 12 | `inputs_ok` | `all` over both readings | gates the whole cycle (§11.1 rule 1) | `[:1]` +4 | RED (test_inverter.py) |
+| 13 | `units` | `tuple` of names | the arbiter's scope table | `[:1]` +4 | RED (5 files) |
+| 14 | `status_payload` | `min(_dues)` over `self.maints` | the operator's "maintenance due" row | `_dues[:1]` +4 | RED (test_maintenance.py) |
+| 15 | `maint_unit` | `self.inverters[self.maint_index]` | which pack the FSM drives | `[0]` −15 | RED (test_maintenance.py) |
+| 16 | store migration | the roster name list | MAINT-05's key migration | `[:1]` +4 | RED (test_maintenance.py) |
+| 17 | `validate_unit_ranges` | the roster argument | startup refusal | `[:1]` +4 | RED (test_inverter.py) |
+| 18 | `cycle()` stale-input hold | `any(sign_wd.tripped)` | **a latched unit's failsafe SET** | `[:1]` +4 | **was GREEN → now RED** (test_watchdog.py part 8) |
+| 19 | `publish()` status | `any(sign_wd.tripped)` | the MQTT status vocabulary | `[:1]` +4 | **was GREEN → now RED** (part 8) |
+| 20 | `status_payload()` banner | `any(sign_wd.tripped)` | the one headline on the page | `[:1]` +4 | **was GREEN → now RED** (part 8) |
+| 21 | `_grade_store_migration` | `any(sign_wd.tripped)` | SAFETY-02's re-arm guard | `[:1]` +4 | was GREEN → now RED (12j), see below |
+| 22 | `cycle()` undeliverable list | `sorted(...)` over shares | the *text* of one event line | `[:1]` +4 | **STILL GREEN — honest gap, below** |
+
+**Rows 18-20 are the finding, and they are not this card's code.** All three are
+`any(i.sign_wd.tripped for i in self.inverters)`, all three are SAFETY-30's, and
+all three survived a green 19-file suite for exactly the reason every blocker on
+this card survived: *the only fixture that reached them was a roster of ONE,
+where `self.inverters[:1]` IS `self.inverters`.* Part 3 of test_watchdog.py has
+built a two-inverter roster with only inverter B latched since SAFETY-30
+shipped — the precise data all three needed — and none of them read it.
+
+Row 18 is the one with teeth. A latched unit must take the whole roster past the
+stale-input hold, because that unit's `failsafe` SET is the thing parking it;
+main.py's own comment says *"the trip would stop acting the moment a poll was
+missed"*. With `[:1]`, a latched inverter B plus a one-cycle gap in inverter A's
+SoC returns early, and B's last share stands for the heartbeat to re-emit.
+test_watchdog.py part 8 now drives that with the latch at **each** position, plus
+the no-latch counterfactual that proves the hold really fires.
+
+**Row 21 — graded structurally, and argued to be redundant.** The `[:1]` there
+is now red on 12j's enumeration, but I could not construct a *reachable* state in
+which it changes an answer: `_grade_store_migration()` runs at startup, before
+any cycle, so a unit is tripped in memory only if its own `unit/<name>/sign_tripped`
+key was truthy — which makes `namespaced_latched` True and short-circuits the
+disjunct. I believe it is an equivalent mutant at that call site rather than a
+live hole, and I am recording the argument instead of asserting the conclusion.
+
+**Row 22 — STILL GREEN, and reported with its size.** `", ".join(sorted(u for u
+in resolution.shares if not self._can_write(u)))[:1]` truncates the **names in
+one event-log line**. It is +4 bytes and no check bites. The refusal itself —
+`elif {u for u in resolution.shares if not self._can_write(u)}` — is a different
+expression and *is* graded (test_peakguard.py part 10h-1b), so the site still
+stops; what a `[:1]` costs is an operator being told one name when two units have
+no writer. It is not a roster comprehension (it iterates `resolution.shares`), so
+12j does not cover it and does not pretend to. Closing it needs a three-inverter
+fixture with two writer-less entries, which is ALLOC-01's fixture and not this
+card's. **Filed as open, with its size and its consequence, rather than closed
+quietly or left unmentioned.**
+
+### And the family closed rather than the member
+
+`test_peakguard.py` part 12j enumerates the family from the AST and requires
+every member to take the roster **whole**, so a new reduction over a slice is red
+**on the line it is written**, at every roster length, with no fixture required.
+That check is what would have caught rounds 2, 3 and 4's blockers on the round
+they were written instead of one review round each; it is also what found rows
+18-21.
+
+⚠️ **AND IT WAS STILL TOO NARROW — see §4d.** Round 4's version enumerated the
+26 roster comprehensions inside the `Controller` class, because that is where
+the four known sites were. Round 5's blockers were four reductions in
+`validate.py`, one module over, over a list *derived* from the roster three
+calls away from it. The pin now parses every `.py` under `app/` and finds the
+roster by transitive closure from one option key: **183 reductions across 9
+modules.**
+
+## 4d. ⚠️ ROUND 5 — the same family one MODULE over, and an instrument that discovers its own scope
+
+Round 4 enumerated the family and closed it **inside the `Controller` class**,
+because that is where the four known sites lived. The review found four more
+survivors immediately, in `validate.py`, on a 19/19 green suite:
+
+| site | mutation | delta | suite |
+|---|---|---|---|
+| `validate.py:779` | `min(caps)` → `min(caps[:1])` | +4 B | **GREEN** |
+| `validate.py:779` | `min(caps)` → `min(caps[-1:])` | +5 B | **GREEN** |
+| `validate.py:778` | `caps = [cap for _, cap in caps_ix[:1]]` | +4 B | **GREEN** |
+| `validate.py:893` | `min(caps_ix[-1:], key=lambda p: p[1])` | +5 B | **GREEN** |
+
+**None of them is an equivalent mutant — a safety refusal disappears.** With
+`max_w: 5000` on inverter-a beside `2000` on inverter-b and
+`sign_disagree_w: 3000`, the shipped validator answers one problem —
+*"sign_disagree_w is 3000 W but max_w (inverter-b) is 2000 W, so the SAFETY-02
+sign-disagreement watchdog can never fire"* — and every one of the four answers
+**zero**. The site starts, with a watchdog reporting `armed` on the status page
+that cannot trip on pack B. That is rule 5's own failure mode reached **through**
+rule 5, which is the sentence `test_inverter.py:1657` says the rule exists to
+prevent. Rule 6 loses its refusal the same way.
+
+### Why round 4's pin could not see them, and why the fix is not four more checks
+
+`caps` and `caps_ix` are not comprehensions over `self.inverters`; they are lists
+**derived** from the roster, three function calls away from it, in a module the
+enumeration never opened. Closing them by hand would have produced a fifth
+rejection with the same shape one step over — that is the whole record of this
+card. **A hand-written scope is the same defect as a hand-written fixture**
+(MUTATION.md habit 12): a check whose data cannot reach the failure it names.
+
+### The instrument
+
+`test_peakguard.py` part 12j now parses **every `.py` under `app/`** off disk —
+no module is named anywhere in the check — and finds the roster by transitive
+closure from **one seed**, the `inverters` option key:
+
+* a function whose returns are roster-carrying is **roster-producing**
+  (`roster_entries` → `build_roster` → `unit_caps_indexed` → `unit_caps`);
+* a name assigned one carries it, **and so does a list accumulated in a loop
+  over one** — which is exactly what `caps`, `roster` and `problems` are, and
+  what round 4's rule could not express;
+* an attribute or property answering one carries it (`self.inverters`,
+  `Controller.packs`, `Controller.units`, `live_units()`);
+* **an argument at a call site taints that parameter in the callee**, which is
+  how the pin walks into `peakguard.binding_soc_floor`, `arbiter.resolve`,
+  `allocator.plan`, `circulation.circulating_w`, `maintenance.arbitrate` and
+  `store.migrate_maint_keys` without being told they exist;
+* and **a slice of a roster is still a roster**, so a mutation cannot cut the
+  taint and hide every reduction downstream of itself. An **index** is one unit
+  and is not.
+
+It then requires, at every place a roster-carrying value is reduced or iterated,
+that the whole of it arrives: no `[:1]`, `[-1:]`, `[:2]` or `[0]` between the
+roster and the reduction.
+
+### What it found — and the hand-written list was wrong in BOTH directions
+
+**183 reductions across 9 modules**, out of 23 `.py` files parsed:
+`main.py` 96, `allocator.py` 24, `validate.py` 19, `arbiter.py` 14,
+`circulation.py` 10, `maintenance.py` 10, `peakguard.py` 4, `inverter.py` 3,
+`store.py` 3 — and 14 modules with no roster reduction at all.
+
+The card's own suggested module list named `watchdog.py`, which reduces over no
+roster at all (SAFETY-30's sign watchdogs are strictly per-unit — there is
+nothing to slice), and did **not** name `maintenance.py` or `store.py`, which
+both do. A pin built from that list would have made a vacuous claim about one
+module and no claim at all about two others. That is the argument for computing
+the scope, made with data rather than with an assertion.
+
+The pin **independently rediscovers all four blockers** — it is told one option
+key and nothing else — and finds a **fifth** beside them at
+`validate.py:783`, the `min(caps_ix, key=...)` inside rule 5's message, which no
+review round named.
+
+### The counterfactual, at two depths
+
+* **Per site, all of them.** Every enumerated site's own expression, in its own
+  scope, mutated with each of the four cuts this card has actually been defeated
+  by. **183 sites × 4 cuts = 732** mutated spellings; every one reported. The same run asserts
+  the predicate **accepts** the legitimate wrapped forms the package really uses
+  (`enumerate`, `zip`, `sorted`, `pack_permissions(packs, floor)`) and does not
+  treat a keyed lookup (`weights[u]` in `allocator._fill`) as a slice — which is
+  what four false SLICED reports cost before `shortens()` was written.
+* **End to end, on the real files.** `[:1]` written into `app/validate.py` at
+  each blocker position, the **whole** analysis re-run over the mutated source,
+  and a sliced site required at that exact line — where the clean package
+  reports none. This is the check that would have taken PR #48 red four rounds
+  ago, with no fixture reaching it.
+
+### The sweep, and the number that must not be quoted alone
+
+All 183 sites were mutated to `[:1]`, one at a time, +4 bytes each, and each
+mutant run against the full 19-file suite in `goodwe-addon-ci` (mounted `:ro`,
+`python3 -B`, a fresh rsync per mutant; 0 `.pyc` in the tree before and after —
+a read-only mount cannot write one, so TEST-15 is closed by construction here
+rather than by discipline). **183 of 183 RED.**
+
+⚠️ **On its own that is tautological**: with the pin in the tree every slice is
+red *by the pin*. So all 51 mutants whose first red was `test_peakguard.py` were
+re-run with the pin **neutralised** (`_SLICED` blanked), leaving only the
+behavioural checks:
+
+| | |
+|---|---|
+| **145 of 183** | a behavioural fixture in this repository separates the site from a slice of itself — red with the pin turned off |
+| **38 of 183** | **nothing but this pin sees it.** Structurally closed by this card; behaviourally **UNGRADED**, and listed rather than counted as covered |
+
+The 38: `main.py` 27, `allocator.py` 6, `validate.py` 2, `arbiter.py` 1,
+`circulation.py` 1, `maintenance.py` 1. Grouped by what a `[:1]` there costs:
+
+* **The operator's page — 15, the largest group.** `Controller.checks`
+  (3310, 3311, 3319, 3328, 3330, 3347, 3368, 3370, 3400, 3458) and
+  `status_payload` (3556, 3566, 3578). A slice reports one unit's row where two
+  exist, or "1 of 2 live" reading 1 of 1. Nothing commands differently; what is
+  lost is the evidence a human uses to notice that it should.
+* **The write path's own counts.** `write_setpoint` 1267/1272/1280 — three
+  `len(targets)` guards. The write itself iterates `targets` at 1287 and *is*
+  graded (test_inverter.py); these three are its log arithmetic and refusal
+  count.
+* **The allocator's internals.** `_fill`'s zero-weight last-resort pool (365),
+  `plan`'s `dropped` (526, 653), `free` (584), `frac_order` (591). ALLOC-01
+  drives two and three units through the *weighted* pass; the last-resort pass
+  and the largest-remainder tie-break are reached, but not with enough units for
+  a slice to change the answer.
+* **Startup and FSM plumbing.** `__init__` 235/351, `_only` 831,
+  `_maint_button` 805, `cycle`'s maintenance zip 1505, `send_heartbeats` 3015,
+  `check_output_clamp` 2445/2452, `check_circulating` 2701,
+  `maintenance.arbitrate` 547, `arbiter.site_resolve` 517,
+  `circulation.opposing` 242.
+* **Two refusal-TEXT sites.** `validate._roster_problems`'s `len(entries)` (554)
+  and `refuse()`'s numbering loop (1263) — both shape the words of a refusal
+  that still happens.
+* **`_grade_store_migration` (698)**, already row 21 of §4c, argued there to be
+  an equivalent mutant at that call site. The argument is unchanged and is still
+  an argument, not a proof.
+
+⚠️ **None of the 38 is closed behaviourally by this card, and saying so is the
+point.** Closing sites one at a time is what cost this card four review rounds.
+The pin closes the **family** — a slice at any of them is red on the line it is
+written — and this list is what a shape pin can never promise.
+
+### And the behavioural half, because a shape pin cannot grade a meaning
+
+The pin grades an **identity** ("the whole roster arrives"). `min` → `max` is a
+zero-byte edit and invisible to it, forever. So `test_validate.py` grows a
+**roster SET** for the three rules that reduce over `unit_caps_indexed()`:
+rosters of width 2, 3, 4 and 5 with the one small cap at **every position** —
+14 rosters × 3 rules = 42 runs of the real validator — plus a **computed family
+table** of twelve candidate reductions (`max`, `first`, `last`, the mean, and
+`[:1]`, `[-1:]`, `[:2]`, `[:3]`, `[1:]`, `[:-1]`, `[::2]`) each compared against
+what the shipped validator actually said on the same roster. Only `min` agrees
+everywhere; any member no roster kills is **printed as a named survivor** and
+takes the check red.
+
+### ⚠️ A THIRD RULE WAS READING THE FLAT `max_w`, AND IT IS SAFETY-30's
+
+Found by reading the three branches side by side — **not** by the pin, and said
+that way because it is the shape of what the pin cannot see. Rule 5's sibling
+branch, `unit_sign_disagree_w` (SAFETY-30's per-inverter sign watchdog), was
+graded against `_number(opts, "max_w", 2000.0)` — the **flat** option — three
+lines under a branch that had already stopped doing that. Both directions were
+measured before the fix:
+
+* a roster of two 2000 W units beside a leftover `max_w: 5000` from a
+  single-inverter install, with `unit_sign_disagree_w: 3000`, produced **zero
+  problems**. The site started, the status page said `armed`, and the per-unit
+  watchdog could not trip on **either** unit. Identical sentence to the blocker,
+  reached through the branch beside it;
+* and a roster of two 5000 W units with no flat key was **refused** at 3000 W
+  against a 2000 W *default the operator never wrote*, for a threshold both
+  units can reach.
+
+It takes the same `weakest` the branch above takes, from the same list, and
+names the same unit. At a roster of one `unit_caps()` is `[max_w]`, so the
+shipped single-inverter refusal is the same comparison, the same boundary and
+the same sentence. Both directions are asserted in `test_validate.py`.
+
+### ⚠️ WHAT THE PIN CANNOT SEE — open rows, with sizes
+
+1. **Semantics.** `min` vs `max` vs the mean: zero-byte edits, invisible to any
+   shape pin. Only the behavioural sets above can see them, and they only see
+   the rules they drive.
+2. **Direct index sites that are not reductions.** `self.inverters[0]` in
+   `_only()` and `actuator_of()`, `self.inverters[self.maint_index]` in
+   `maint_unit`, `entries[index]` in `_unit_label()`. Single-unit lookups, meant
+   to be, graded where they live.
+3. **A subscript by a NAME.** `weights[u]`, `cap[u]`, `add[u]` in
+   `allocator._fill` are keyed lookups into a roster-derived mapping and are
+   accepted deliberately, so `caps[i]` with a variable `i` would pass. Not in
+   the family this card has been defeated by; buying it costs four false reports.
+4. **A slice through an intermediate call.** `sorted(roster)[:1]` subscripts a
+   call result, not a roster reference. Nothing in `app/` is written that way
+   today; if something is, it would not be seen.
+5. **Anything outside `app/`.** `tools/`, `sim/` and the checks themselves are
+   not parsed — and `sim/inject.py` reaches `validate_startup_config()`.
+6. **The taint's own seam.** It crosses a call boundary only where the callee is
+   a plain function this package defines and the argument is positional or
+   keyword. A roster handed through `*args`, stored in a dict and read back, or
+   passed to a method on an unresolvable object, is lost at that hop — and
+   everything downstream goes **unenumerated** rather than reported unreachable.
+   That is the shape of the remaining exposure: not zero, and not measurable
+   from inside the check.
+7. **`cycle()`'s undeliverable-unit log line** (row 22 of §4c) is still green at
+   +4 bytes, unchanged from round 4 and still filed open.
+
+## 5. ⚠️ THE RESIDUAL — a per-pack EMS floor is NOT this card
+
+`peak_soc_floor` is still **one number for the site**, so the card's own
+"or the deeper pack is needlessly withheld" half is not closed and this record
+does not pretend it is. With cut-outs of 10 % and 18 %, rule 8 obliges the
+operator to set the EMS floor above 18, and the 10 % pack is then held above its
+own need. Closing that means `peak_soc_floor` becoming a per-entry key too, which
+is a second option migration and a product decision about what the site-wide
+number then means. The card scoped itself to `bms_soc_floor` and explicitly told
+this developer not to widen it, so it is filed here as the open question it is
+rather than smuggled in.
+
+**Also still not in this card, and named by the shipped code already:** the
+charge *ceiling* is the mirror image — `pack_floor_soc` is the wrong reduction
+for it, because a ceiling is the **fullest** pack and `arbitrage`'s `soc_ceiling`
+takes one number. `main.py:932-936` says so, and it belongs to CFG-04/SPIKE-04's
+per-pack bound. Nothing here touched it.
+
+## 6. What was changed, and one thing that was found on the way
+
+* `config.yaml` — `bms_soc_floor: int(0,50)` required per `inverters:` entry;
+  the flat key and its default are untouched and are documented as the
+  single-inverter shape's own cut-out.
+* `app/inverter.py` — `Inverter.bms_soc_floor`; `_entry_bms_floor()` (`None`, not
+  a borrowed number and not `0.0`); `_flat_bms_floor()`.
+* `app/peakguard.py` — `Pack`, `pack_floor_pct()`, `pack_permissions()`,
+  `binding_soc_floor()`. `discharge_permitted()` is unchanged and is still the
+  one place the question is answered.
+* `app/validate.py` — `unit_bms_floors_indexed()`; rule 8 grades every unit and
+  names it via the **roster** index the value came off.
+* `app/main.py` — `Controller.packs`; the `decide(soc_floor=…)` reduction; the
+  floor-limited notification names the pack(s) at or below their own floor, at a
+  roster of two or more only.
+* `tools/supervisor_shape_probe.py` — **re-run, and repaired.** MUTATION.md
+  habit 8: a committed transcript is a check and it decays. Making the key
+  required broke §C outright and, worse, left §D printing **PASS** on a refusal
+  about the *wrong key* — its "no `max_w`" case was answered `Missing option
+  'bms_soc_floor'`. §D is now an isolated counterfactual over a **merged,
+  accepted** baseline with exactly one key varied per case, asserting the key's
+  **name** in the refusal, and it carries a `bms_soc_floor` case in both
+  directions. Re-run against `ghcr.io/home-assistant/amd64-hassio-supervisor:2026.08.0`:
+  `all cases passed`.
+* `app/inverter.py`, found by the mutation battery — `_flat_bms_floor()` was
+  written with the default in **two** places (the `.get()` fallback *and* the
+  blank test). Deleting either one survived, because each covered for the other:
+  MUTATION.md's masking section in three lines of option reading. It has one
+  defaulting line now, and M12 takes six checks red.
+
+* **Round 3 changed no `app/` file at all.** `test_peakguard.py` only:
+  parts 12c2 and 12g. Its battery was **46 mutations across
+  `binding_soc_floor()`'s reduction, the `decide()` call site, `Controller.packs`
+  and `pack_floor_pct()`'s conjunction** — bytecode purged and counted before and
+  after each (23 → 0). **43 red, 3 survivors, all three recorded above and in the
+  in-process rows**: `max(floors[:12])` and `max(floors[-12:])`, which need a
+  13-pack roster to observe and are reported with that size rather than hidden;
+  and `max(set(floors))`, which is **equivalent** — `max` is duplicate-invariant,
+  so it cannot change an answer. `list(reversed(self.packs))` is equivalent too
+  (`max` is order-invariant) and is nonetheless red on the structural pin.
+
+⚠️ **No measurement of inverter B appears anywhere in this change.** It does not
+physically exist. Every second-pack number in a fixture, in the probe and in
+config.yaml's commented example is **simulated and labelled as such**; only the
+10 % on inverter A is read off hardware. The three-pack runs added in round 3 are simulated
+twice over: there is no inverter C at all, and the shallow slots carry inverter
+A's measured 10 % while the 35 % cut-out that moves along the roster is invented.
+
+---
+
+## 7. The probe transcript, re-run 2026-09-05 on 2026.08.0
+
+Command (unchanged, in `tools/supervisor_shape_probe.py`'s own docstring):
+
+```
+    docker run --rm -v "$PWD:/w:ro" -v "$PWD/tools:/probe:ro" \
+      --entrypoint python3 \
+      ghcr.io/home-assistant/amd64-hassio-supervisor:2026.08.0 \
+      /probe/supervisor_shape_probe.py
+```
+
+It reads `config.yaml` and runs Supervisor's own `AppOptions` validator in a
+throwaway container. It touches no live Supervisor, POSTs nothing and installs
+nothing.
+
+⚠️ **RE-RUN AGAIN ON 2026-09-05**, because the rebase onto `67e649a` merged
+SPIKE-04a's `max_w: 4600` into the same §C/§D blocks this card had rewritten,
+and MUTATION.md habit 8 says a committed transcript that a document quotes is a
+check. A hand-merged transcript is not a measurement of anything. The block
+below is the output of that run, not of the 2026-09-04 one: B's rating reads
+**4600** — its datasheet figure, still not a measurement of any unit — and every
+case still passes.
+
+```
+== C. AC2: a two-inverter blob an operator would POST ==
+  PASS  two inverters
+        [{"name": "inverter-a", "setpoint_entity": "number.goodwe_a_setpoint_w", "soc_entity": "sensor.goodwe_a_soc", "batt_entity": "sensor.goodwe_a_battery_power", "max_w": 5000, "bms_soc_floor": 10}, {"name": "inverter-b", "setpoint_entity": "number.goodwe_b_setpoint_w", "soc_entity": "sensor.goodwe_b_so
+
+== D. what the schema itself refuses, per entry ==
+  PASS  the per-entry BASELINE is ACCEPTED, so every refusal below has nowhere else to come from
+        [{"name": "inverter-b", "setpoint_entity": "number.b", "max_w": 4600, "bms_soc_floor": 10}]
+  PASS  an entry with no name is refused, and the refusal NAMES it
+        Missing option 'name' in inverters in GoodWe RS485 Controller (local_goodwe_controller)
+  PASS  an entry with no setpoint_entity is refused, and the refusal NAMES it
+        Missing option 'setpoint_entity' in inverters in GoodWe RS485 Controller (local_goodwe_controller)
+  PASS  an entry with no max_w is refused, and the refusal NAMES it
+        Missing option 'max_w' in inverters in GoodWe RS485 Controller (local_goodwe_controller)
+  PASS  an entry with no bms_soc_floor (SAFETY-32) is refused, and the refusal NAMES it
+        Missing option 'bms_soc_floor' in inverters in GoodWe RS485 Controller (local_goodwe_controller)
+  PASS  an entry with max_w over the range is refused, and the refusal NAMES it
+        value must be at most 5000.0
+  PASS  an entry with bms_soc_floor over the range (SAFETY-32) is refused, and the refusal NAMES it
+        value must be at most 50.0
+  PASS  a JSON STRING where the list belongs is refused
+        Invalid list for option 'inverters' in GoodWe RS485 Controller (local_goodwe_controller)
+
+== E. the ROLLBACK direction, and the operator's UI ==
+  PASS  a schema WITHOUT `inverters` drops the key with a warning rather than refusing - so a downgrade is not the §B trap in reverse
+        ['arbitrage_enabled', 'arbitrage_soc_ceiling', 'arbitrage_soc_floor', 'auto_start', 'batt_entity', 'batt_invert']
+  PASS  Supervisor's UI renderer turns it into a repeatable section rather than choking - an operator can add the second inverter in the Configuration tab, which the JSON-string alternative could not offer
+        [{"name": "inverters", "type": "schema", "optional": true, "multiple": true, "schema": [{"name": "name", "required": true, "type": "string"}, {"name": "setpoint_entity", "required": true, "type": "string"}, {"name": "max_w", "lengthMin": 100.0, "lengthMax": 5000.0, "required": true, "type": "integer
+
+== all cases passed ==
+```

goodwe_controller/DOCS.md

@@ -78,9 +78,20 @@ inverters:
     setpoint_entity: number.goodwe_b_setpoint_w
     soc_entity: sensor.goodwe_b_soc
     batt_entity: sensor.goodwe_b_battery_power
-    max_w: 3000
+    max_w: 4600
 ```
 
+⚠️ **The 4600 W is inverter B's *datasheet* figure, not a measured one.** B is a
+GoodWe GW5048D-ES (serial `55048ESU22AW0992`, identified 2026-09-04) whose
+datasheet gives max charging and discharging power as 4600 W; with nothing
+connected to its PV inputs the battery path is the ceiling, not the 5000 VA AC
+nameplate. The datasheet prints that number under *Battery Input Data*, so it
+may be a DC figure, while `max_w` is graded against register `0x0518` — the
+inverter's own **AC** power. If it is DC the AC-side ceiling is nearer 4324 W.
+`max_w: 4600` is still the right thing to configure, because the freeze rail is
+`min(max_w, dev_max)` and the device range read off the unit is what binds.
+`goodwe/NOTES.md` §1.1b records both readings and picks neither.
+
 | per-entry key | required | meaning |
 |---|---|---|
 | `name` | yes | Short identifier. It namespaces this unit's SAFETY-02 latch, heartbeat and write bookkeeping in the add-on's state file, so **two units may not share one** |
@@ -363,6 +374,10 @@ behind it would trip the firmware watchdog on a system that is working fine.
 | `auto_start` | false | Start controlling on boot (only after commissioning) |
 | `sign_disagree_w` | 300 | Both the command and the meter must exceed this before a sign difference counts. Must stay **below `max_w`** or the watchdog can never fire, and the add-on refuses to start |
 | `sign_disagree_s` | 180 | How long they must stay opposed, continuously, before the failsafe trips |
+| `unit_sign_disagree_w` | 250 | **Two or more inverters only.** Both a unit's own share and its own measured power must exceed this before their sign difference counts. It is `saturation_w / 2`, so the smallest divergence it can see is exactly the gap at which the loop already calls the inverter "not following". Must stay **below `max_w`** or it can never fire, and the add-on refuses to start |
+| `unit_sign_disagree_s` | 85 | **Two or more inverters only.** How long one unit's command and its own power must stay opposed before that unit is parked |
+| `site_sign_disagree_w` | 300 | **Two or more inverters only.** Threshold for the site-wide check - the sum of the commands against the meter. The same 300 W as `sign_disagree_w`, for the same reason: near zero the sign is noise |
+| `site_sign_disagree_s` | 325 | **Two or more inverters only.** How long the site's total command and the meter must stay opposed before every inverter is parked. Longer than `sign_disagree_s` because the command has to cross zero over a rail that is the sum of the roster, at a `slew_w` that is not - measured at 110 s with one inverter and 215 s with two |
 
 #### The sign-disagreement failsafe
 
@@ -374,6 +389,43 @@ there. The banner reads `FAILSAFE`, the commissioning list says which two
 numbers disagreed and for how long, and the `Controller status` entity reports
 `failsafe`.
 
+#### ⚠️ At two or more inverters this is TWO checks, not one (SAFETY-30)
+
+The paragraph above describes what a **one-inverter** install does, and it is
+unchanged: there is one meter, one actuator, and comparing the command against
+the meter is comparing a thing with a view of itself.
+
+With a second inverter that stops being true. The meter answers for **both**
+units, so a command put on one of them can be opposed to it with nothing wrong
+(inverter A charging while inverter B discharges more), and - the worse
+direction - a unit genuinely driving the wrong way can be **hidden**, because its
+sibling's watts arrive at the meter and the sum looks healthy. So the check
+splits in two, and they are different faults:
+
+- **Per inverter** — that unit's own share against **its own measured power**
+  (the `batt_entity` reading, register 0x0518). This catches *"this inverter is
+  not doing what it was told"*. Only that inverter is parked; the site keeps
+  running on the one that works. Thresholds: `unit_sign_disagree_w` /
+  `unit_sign_disagree_s`.
+- **Site-wide** — the **sum** of what every inverter has standing, against the
+  meter. This catches *"the site as a whole is pushing the wrong way"*, which no
+  per-inverter check can see because every unit may be faithfully executing its
+  share of a number that is itself wrong. **Every** inverter is parked and the
+  banner names the site rather than an inverter. Thresholds:
+  `site_sign_disagree_w` / `site_sign_disagree_s`, latched under flat
+  `site_sign_tripped` in `/data/state.json`.
+
+Both latch, both are cleared by the same *stop, then start*, and both survive a
+restart. **On a one-inverter install neither pair does anything at all** - the
+two questions are the same question there, and only the check described above is
+armed.
+
+This is **not** the circulating-power alarm further down: that one compares the
+inverters *against each other* and catches them pushing power through each other,
+which nets to zero at the meter and which no command is evidence about. It is
+also not the saturation freeze, which is a *magnitude* test that shapes the loop
+rather than a *direction* test that parks an inverter.
+
 **It does not clear itself, and restarting the add-on does not clear it either.**
 Pressing *Start / stop control* twice — stop, then start — is the only thing that
 does. That is deliberate on both counts:
@@ -513,7 +565,7 @@ help - see "Why the tuning is what it is".
 | `peak_forecast_entity` | Quarter-hour demand forecast, for capacity-tariff markets. Empty = no cap |
 | `peak_cap_w` | The site's capacity-tariff target |
 | `peak_soc_floor` (20) | **The percent below which the capacity-peak guard will not discharge the battery.** ⚠️ It bounds **the peak guard only** — it is **not a pack-wide reserve**, and ordinary self-consumption discharges straight past it (SAFETY-25, see below). Not optional and not the same thing as `maintenance_soc_floor` |
-| `bms_soc_floor` (10) | The pack's **own** cut-out percent, the number the BMS enforces whatever you set above. `peak_soc_floor` must be **strictly above** it or the add-on refuses to start |
+| `bms_soc_floor` (10) | The pack's **own** cut-out percent, the number the BMS enforces whatever you set above. `peak_soc_floor` must be **strictly above** it or the add-on refuses to start. ⚠️ **At a declared `inverters:` roster this key moves INTO each entry and is REQUIRED there** (SAFETY-32): it describes a battery, and two packs of different models have different cut-outs, so one number for both protects the shallower pack with the deeper pack's BMS. Nothing is inherited — an entry that omits it is refused **by name** at startup, and the flat key below is then not read at all |
 | `price_now_entity`, `price_avg_entity` | Dynamic tariff. Empty = never force a paid grid top-up |
 
 On a capacity-tariff site the maintenance charge is capped by the headroom left

goodwe_controller/MUTATION.md

@@ -135,11 +135,19 @@ manufactures them from a **fixture**: the check runs, the assertion is
 evaluated, the sentence above it is true — and the numbers it compares are
 arranged so that the failure it describes could not have changed the answer.
 
-It is not hypothetical and it is not rare. Six instances in two weeks; three
+It is not hypothetical and it is not rare. EIGHT instances in two weeks; four
 were found by review, two - CFG-04's fixture pair - by an author's own mutation
-battery, which is the cheaper place, and the sixth by review again, in the one
+battery, which is the cheaper place, one by review again, in the one
 place nobody had thought to point the rule at: not a check in the suite, but a
-COMMITTED EVIDENCE ARTIFACT four documents cite.
+COMMITTED EVIDENCE ARTIFACT four documents cite - and the eighth INSIDE THE FIX
+WRITTEN TO CLOSE THE SEVENTH.
+
+⚠️ **AND IT HAS NOW APPEARED IN SIX CONSECUTIVE CARDS.** The seventh instance,
+MAINT-05's, is the one that says why: the defect is not carelessness about
+fixtures, it is that **a check and the thing it checks were built from the same
+constant**. Ask of every difference-claim not "is this assertion true" but
+"where did each side of this comparison come from" - and if the answer is the
+same place twice, the check is decoration.
 
 * **ARB-03** shipped a call-site scan asserting that every `claims.append` in
   `cycle()` routes through `actuator_of()`. Its own parse **saw 3 of the 8
@@ -167,6 +175,44 @@ COMMITTED EVIDENCE ARTIFACT four documents cite.
   than against its behaviour. It now drives the shipped `run_control()` on a
   two-inverter config, **under a deadline**, because the mutation's real symptom
   is that the controller enters its loop and never returns.
+* **MAINT-05** shipped `Controller.__init__` calling
+  `migrate_maint_keys(store, self.inverters[0].name, ...)` under a comment
+  saying *"it targets `self.inverters[0].name` and NOT `UNIT_A` ... writing
+  `UNIT_A` here would deposit an install's whole maintenance history under a
+  name that is not on the roster at all"*. `inverter.UNIT_A` **is** the string
+  `"inverter-a"`, and every two-pack fixture in `test_maintenance.py` named its
+  first roster entry `inverter-a` — so the two expressions were the same value
+  in every check that existed, and swapping them left **all 19 files green**
+  while losing a real install's whole maintenance history. Two-pack fixtures are
+  now `["garage", "shed"]`, each behind a precondition check (habit 1) that the
+  first name is not `UNIT_A`.
+
+  ⚠️ **AND THE POSTCONDITION HAD THE SAME SHAPE AS THE FIXTURE.**
+  `store._verify_moves()` read back the keys it had just written, under the
+  namespace it had just been handed — so it verified clean, and the caller
+  stamped the schema marker, on a migration nothing would ever read. *A check
+  that reads back the namespace it was handed cannot detect being handed the
+  wrong one*: it compared a thing against itself, exactly as DEV-02's fixture
+  compared `(-5000, 5000)` against `(-5000, 5000)`. The habit-2 fix is the same
+  one — grade against something the thing under test did not supply, here the
+  ROSTER the machines read.
+
+  ⚠️ **AND THE FIX ITSELF CARRIED THE DEFECT, WHICH IS THE SIXTH CARD IN A ROW
+  AND THE FIRST TIME IT HAS APPEARED *INSIDE* A REMEDY.** The fix round's
+  recovery check was `migrate_maint_keys(_m5, ROSTER[0], ROSTER)` — the
+  **corrected** name. It measures recovery along the one path where the operator
+  has *already fixed the bug*, and nothing in the 19 files called
+  `migrate_maint_keys()` twice with the **same wrong** `unit`. So the refusal
+  the card was rejected twice to obtain lasted **exactly one start**: measured
+  at Controller level over one persisted store, roster `["garage", "shed"]` —
+  construction 1 `ok=False misdirected=4 moved=4 kept=0 marker=<absent>`,
+  construction 2 `ok=True misdirected=0 moved=0 kept=4 marker=1`. On start 2
+  rule 1 routes the wrongly-written keys to `rep.kept`, `rep.moved` is empty,
+  the roster loop iterates over nothing, and the marker is stamped on a
+  migration that reached no readable key — after which the early return skips it
+  forever, *including after the call site is fixed*. The check was true and
+  passed for the wrong reason. Habit **9** below is what falls out of it.
+
 * **CFG-04's `tools/supervisor_shape_probe.py` §B** — and this one is the rule
   reaching outside the suite. The probe is the committed evidence that
   `config.yaml`'s `inverters: []` default is load-bearing, and four documents
@@ -241,6 +287,81 @@ Cheap habits fall out of it, and all of them are now used in this repository:
    red. `inverters: []` was deleted from `config.yaml` and the probe re-run
    against `ghcr.io/home-assistant/amd64-hassio-supervisor:2026.08.0` for
    exactly that reason.
+9. **Exercise a refusal a SECOND time with the SAME wrong input — not the
+   corrected one.** A guard that refuses once and then admits the same bad
+   caller is not a guard, it is a delay, and the second call is a *different
+   code path* whenever the first call left state behind. MAINT-05's roster
+   grading iterated the keys it had just **written**; on the second start the
+   same wrong caller found them already present, they moved to the
+   `already namespaced` list, and the loop graded an empty set. Re-running with
+   the **fixed** argument — which is what the check did — measures the operator's
+   recovery, never the persistence of the refusal. Ask of every refusal: *what
+   does this do when the caller has not been fixed and simply runs again?* And
+   assert the **shape** of that second report, not only its verdict, so the
+   check cannot silently revert to grading the first path.
+
+10. **A fixture of N members cannot separate two reductions that agree on N
+    members — so grade the FAMILY, and grade it from the SOURCE.** This is the
+    only lesson SAFETY-32 cost four review rounds to learn, and it generalises
+    past packs to every roster, list and window in this repository. Two packs
+    cannot tell `max` from `min` from first from last. Three cannot tell `max`
+    from `max(x[-2:])`. **The regress does not terminate in a fixture**: `[:1]`
+    needs two members, `[:2]` three, `[:3]` four, and adding one more member
+    moves the hole rather than closing it. Three things follow, and all three
+    are now in `test_peakguard.py` parts 12c2, 12g, 12i and 12j:
+    * the fixture is a **SET** — the extreme at every position of the roster,
+      not one roster with the extreme somewhere;
+    * the family table is **COMPUTED and PRINTED**, so a member no fixture kills
+      shows up as a named survivor and takes the check red, instead of being
+      absent from a list nobody re-derived;
+    * where the claim is an **IDENTITY** ("this roster", "this list"), pin it
+      **exactly on the AST** — `ast.unparse(arg) == "self.packs"` is false of
+      every slice, where `"self.packs" in src` is true of all of them. Where the
+      claim is **SEMANTIC** ("the maximum"), a shape pin would refuse honest
+      refactors, so report the unreached width **with its size** instead.
+
+11. **When a check of this shape is found, mutate its SIBLINGS before shipping
+    the fix.** SAFETY-32 was rejected three times for the same defect one step
+    over: the reduction, then the roster handed to it, then the reduction handed
+    to the *other argument of the same call*. Enumerating the family from the
+    AST and mutating each member took one afternoon and found **four more
+    survivors in code the card never touched** — three of SAFETY-30's
+    `any(i.sign_wd.tripped for i in self.inverters)` reductions, invisible
+    because the only fixture that reached them was a roster of ONE, where
+    `self.inverters[:1]` **is** `self.inverters`. Ask of every fix: *what is the
+    sibling of the thing I just hardened, and does any fixture separate it from
+    a slice of itself?*
+
+12. **When the same shape has been found twice, STOP CLOSING SITES AND BUILD
+    THE INSTRUMENT — and make it DISCOVER its own scope.** SAFETY-32 was
+    rejected four times, and each round closed exactly the site it was handed:
+    the reduction, then the roster handed to it, then the *sibling argument* of
+    the same call, then the same family one *module* over, in validate.py.
+    Round 4 did enumerate — and enumerated inside the `Controller` class,
+    because that is where the four known sites lived. **A hand-written scope is
+    the same defect as a hand-written fixture**: it is a check whose data cannot
+    reach the failure it names, and it goes on passing while the thing it was
+    supposed to cover sits in the file nobody added to the list. Round 5's pin
+    parses every `.py` under `app/`, seeds itself from ONE option key, and
+    propagates through returns, accumulator loops, attributes and *call
+    arguments* to find the roster wherever it went — 183 reductions across 9
+    modules. It rediscovered all four blockers without being told where to look,
+    and it named two modules the hand-written list would not have contained
+    (`maintenance.py`, `store.py`) while *omitting* one the list did contain
+    (`watchdog.py`, which is strictly per-unit and reduces over no roster at
+    all). **The list was wrong in both directions**, which is exactly what a
+    computed scope is for.
+
+13. **Say what the instrument CANNOT see, with sizes, in the same breath.** A
+    structural pin grades an **identity** ("this whole roster arrives"); it can
+    never grade a **meaning** (`min` vs `max` is a zero-byte edit and invisible
+    to it). Ship the pin AND the behavioural set, name the gap, and give every
+    unreached case its byte delta — `cycle()`'s undeliverable-unit log line
+    (+4 B, still green), a subscript by a *variable* rather than a constant, a
+    slice taken through an intermediate call, anything outside `app/`. **An
+    enumerated list with an honest gap is worth more than a claim of
+    completeness a reviewer falsifies in ten minutes**, and every round of this
+    card that claimed the latter was rejected for it.
 
 And the mutation-table consequence: **a variant is not covered because a
 near-miss of it is.** DEV-02's table listed the *actuator* variant of the
@@ -297,6 +418,8 @@ commit and PR bodies.
 | **TARIFF-01** reviewer survivor: `record_block()` ignoring its `policy` argument (0dbd485) | **re-verified — still a survivor** | re-run here through `tools/mutation_run.sh` against the 15 suites that were on the record, all green with `billed_w(block, policy)` mutated to `billed_w(block, POLICY_NET)`. The recorded diagnosis holds: every block reaching the store is single-phase, where NET, SUM and MAX collapse. **Now CLOSED by TEST-13** (`test_tariff.py` section 7), which round-trips section 7's own unbalanced `(3000, 500, -200) W` block THROUGH the store under each policy - three distinct banked figures, so a hardcoded policy takes at least two of them red. Third instance of "a stimulus evaluated where the quantity it grades is degenerate", after WITNESS-02's span normalisation and TARIFF-01's own DST window. |
 | **SAFETY-25** 8 mutations against `test_peakguard.py` part 10, then 9 more after review, all red | **not at risk** | each run against a tree copied fresh with `__pycache__` excluded and mounted `:ro`, so no `.pyc` existed to go stale - the same structural immunity recorded for CI above. Byte deltas recorded on the PR; none is `+0 B`. All are **reds**, which is the direction this defect cannot fabricate. ⚠️ The second round exists because the first round's mutations were all aimed at DISARMING the guard, and the claim part 10 actually made was about the guard being FIXED - the reviewer built the fix and the suite stayed green. Bytecode was never the issue there; an unaimed battery was. |
 | **SAFETY-03** reviewer survivor: `abs()` dropped from `clamp_to_rating`'s limit (6aabb83) | **re-verified — still a survivor** | re-run the same way, all 15 green. Also as recorded: unreachable under the `int(100, 5000)` schema. |
+| **MAINT-05 round 3**, `migrate_maint_keys()` marker gate `if rep.ok:` → `if rep.ok or rep.kept:` (`+12 B`) | **was a survivor — now CLOSED** | measured against head `65cddf3` under the purge, tree mounted `:ro`, `19/19 PASS`. It stamps the schema marker on a migration that LOST state, defeating the older `lost` guard as well as the new `misdirected` one, because every check that reached the marker did so on a store whose `kept` list was empty. Closed by `test_maintenance.py` **9d**; the mutation is now red on three named checks. |
+| **MAINT-05 round 3**, `migrate_unit_keys()` marker gate `if rep.ok:` → `if rep.ok or rep.kept:` (`+12 B`) | **found here — SAFETY-02's half, now CLOSED** | the SAME mutation one function up, on the LATCH migration rather than the history one. Measured a survivor against this branch under the purge, `:ro`, `19/19 PASS`, then closed by `test_inverter.py` part 10's new fixture — a store with `unit/inverter-a/sign_reason` already present (so `kept` is non-empty) whose `sign_tripped: true` write is dropped. Shipped: marker withheld. Mutant: marker stamped on a **disarmed latch**, and the next start skips the migration that would have re-armed it. |
 | **TARIFF-01**'s other three survivors, **SAFETY-03**'s `describe()` survivor, **TARIFF-02**'s M3/M5/M7, **SAFETY-18**'s three | **still unverified** | recorded before the purge discipline, with no byte delta on the record, so a `+0 B` same-second revert cannot be excluded. All are survivors, which is the direction this defect fabricates. TARIFF-02's and SAFETY-18's were **closed by later commits whose claims are reds**, so what is unverified is the *diagnosis*, not the current state of the code. |
 
 **Nothing here says a past result is wrong.** The two survivors re-run above

goodwe_controller/app/arbiter.py

@@ -151,6 +151,30 @@ CLAIM_SCOPE = {
     # other went on feeding the loop. The evidence is site-level and so is the
     # remedy: stop commanding, everywhere, and say why.
     "circulation": SCOPE_SITE,
+    # main.py: `Claim.set("sign_site", P_SAFETY, 0.0, "site sign disagreement")`.
+    # SAFETY-30, and it is the twin of `failsafe` below rather than a duplicate
+    # of it - the two exist BECAUSE the same words describe two different
+    # measurements once there is more than one inverter.
+    #
+    # `failsafe` grades ONE unit's standing command against THAT UNIT'S OWN
+    # measured power (register 0x0518). Its conclusion is "the actuator I am
+    # driving is not doing what I told it", which is a fact about one device,
+    # so it parks one device.
+    #
+    # This one grades the SUM of the standing commands against P1. Its
+    # conclusion is "the site as a whole is driving the wrong way" - an inverted
+    # gain, an integrator wound past the rail, a strategy on the wrong side of
+    # the meter - and there is nothing per-unit about it: the quantity is a sum
+    # over the roster and the offending number is the site target, which no
+    # single inverter owns. Parking one unit on that evidence would leave the
+    # rest of the site pushing the same way with the guard satisfied.
+    #
+    # ⚠️ AT N=1 IT IS NEVER FILED, and that is the identity rather than a
+    # special case: with one inverter the sum of the commands IS that unit's
+    # command and `failsafe` already carries exactly this trip, under the keys,
+    # the banner and the claim an install has had since SAFETY-02 shipped.
+    # `Controller.cycle()` does not even arm the site detector there.
+    "sign_site": SCOPE_SITE,
 
     # --- UNIT: about one pack, one inverter, one device -------------------
     # main.py: `Claim.limit("device", P_SAFETY, self.dev_min, self.dev_max,
@@ -164,13 +188,22 @@ CLAIM_SCOPE = {
     # same reason, so it is scoped like `device`.
     "supervised": SCOPE_UNIT,
     # main.py: `Claim.set("failsafe", P_SAFETY, 0.0, "sign disagreement")`.
-    # ⚠️ THE EVIDENCE IS SITE-WIDE AND THE REMEDY IS NOT, and that is why this
-    # entry needs its reasoning written down. SAFETY-02 compares P1 - a site
-    # measurement - against `standing`, THE COMMAND THIS CONTROLLER PUT ON THIS
-    # ACTUATOR. What it concludes is "the unit I am driving is pushing power the
-    # wrong way", and what it does is park that unit. At N>=2 the same evidence
-    # cannot be attributed to a sibling nobody here commands, so a site-wide
-    # remedy would be this watchdog stopping an inverter it never graded.
+    # ⚠️ THE EVIDENCE IS PER-UNIT AND SO IS THE REMEDY - AND SINCE SAFETY-30 THE
+    # FIRST HALF OF THAT SENTENCE IS TRUE. It used to read "the evidence is
+    # site-wide and the remedy is not", because the watchdog compared THIS
+    # unit's standing command against P1: a site measurement, attributed to one
+    # actuator on the strength of §6.2's N=1 identity. That identity does not
+    # survive a second inverter - at N>=2 the meter answers for a sibling this
+    # unit's command says nothing about, in both directions (a false trip when
+    # the sibling covers the meter, and a missed one when the sibling MASKS a
+    # unit that really is reversed).
+    #
+    # So at N>=2 the comparison behind this claim is THIS unit's command against
+    # THIS unit's own measured power - register 0x0518, per-instance since
+    # DEV-02 - and the scope that was already right for the wrong reason is now
+    # right for the right one. The site-wide half moved to `sign_site` above.
+    # At N=1 the two are the same comparison and this claim is the one filed,
+    # exactly as it shipped.
     "failsafe": SCOPE_UNIT,
     # main.py: `Claim.set("maintenance", P_MAINTENANCE, ...)` when the cycle owns
     # the setpoint, and `Claim.limit("maintenance", P_MAINT_SHAPE, hi=...)` when

goodwe_controller/app/inverter.py

@@ -15,6 +15,11 @@ The scalars, and what each one was silently asserting:
                           SAFETY-24's pack floor and arbitrage's bounds. An
                           average lets the emptier pack be driven below its own
                           floor; a maximum parks the fuller one.
+  `bms_soc_floor`         SAFETY-32. ONE cut-out percent for every pack, and it
+                          describes A BATTERY. Two packs of different models
+                          have different cut-outs, so the flat key graded the
+                          shallower pack against the deeper one's BMS - wrong
+                          in the direction that empties a pack.
   `batt`                  ONE battery power into the anti-windup saturation
                           detector - see the SITE/UNIT note at the bottom of
                           this docstring, because this one goes the OTHER WAY.
@@ -117,6 +122,12 @@ import math
 from datetime import datetime
 
 from .clamp import ClampResult
+# CFG-03's ONE in-code mirror of config.yaml's SAFETY-24 defaults. Imported
+# rather than retyped: the flat-keys path below needs `bms_soc_floor`'s
+# shipped default and a literal here would be the fourth defaults site that
+# ticket exists to prevent. peakguard imports arbiter/control/tariff and none
+# of those imports this module, so the direction is one-way.
+from .peakguard import PEAK_OPTION_DEFAULTS
 from .store import unit_key
 from .watchdog import SignWatchdog, SignWatchdogConfig
 
@@ -170,7 +181,7 @@ class Inverter:
 
     def __init__(self, name, *, setpoint_entity="", soc_entity="",
                  batt_entity="", batt_invert=False, max_w=2000.0,
-                 store=None, sign_cfg=None):
+                 bms_soc_floor=None, store=None, sign_cfg=None):
         self.name = str(name)
         self.setpoint_entity = str(setpoint_entity or "")
         self.soc_entity = str(soc_entity or "")
@@ -179,6 +190,19 @@ class Inverter:
         # The supervised cap on what may be written to THIS setpoint entity.
         # See the module docstring on why this is a number and not a `Tuning`.
         self.max_w = float(max_w)
+        # SAFETY-32. THIS PACK'S OWN BMS CUT-OUT, in percent, or None when the
+        # roster entry did not state one.
+        #
+        # ⚠️ NOT DEFAULTED HERE, AND None RATHER THAN A NUMBER. A cut-out is a
+        # property of the battery in the cupboard; inventing one for a pack that
+        # did not state it is precisely this card's defect - a pack graded
+        # against a floor that belongs to a different battery - rebuilt in the
+        # constructor. `peakguard.pack_floor_pct()` treats None as an unknown
+        # floor and `discharge_permitted()` treats an unknown floor as NO
+        # permission, so the absence costs a shave; validate.py rule 8 refuses
+        # the config by name long before that, at startup.
+        self.bms_soc_floor = (None if bms_soc_floor is None
+                              else float(bms_soc_floor))
 
         # --- the device's own reported range, read from its own entity -------
         # None until `hass.limits()` has answered. NOT guessed at: ESPHome is
@@ -339,8 +363,9 @@ def build_roster(opts: dict, store, sign_cfg=None) -> list:
     TWO SHAPES, AND WHICH ONE IS IN FORCE IS DECIDED BY ONE TEST: is
     `inverters:` empty?
 
-      EMPTY   the five flat keys - `setpoint_entity`, `soc_entity`,
-              `batt_entity`, `batt_invert`, `max_w` - describe the ONE inverter,
+      EMPTY   the six flat keys - `setpoint_entity`, `soc_entity`,
+              `batt_entity`, `batt_invert`, `max_w`, `bms_soc_floor` (SAFETY-32)
+              - describe the ONE inverter,
               exactly as the shipped scalars did. This is what every install
               that upgrades into this version reads, without an operator
               touching anything, and it is byte-identical to what DEV-02
@@ -353,8 +378,8 @@ def build_roster(opts: dict, store, sign_cfg=None) -> list:
               `soc_entity` has no SoC sensor; it does NOT borrow the flat one,
               because "this unit's number came from somewhere else" is the
               defect DEV-02 spent a card removing and a fallback here would
-              rebuild it in three lines. `name`, `setpoint_entity` and `max_w`
-              are required of every entry - the schema enforces it on the
+              rebuild it in three lines. `name`, `setpoint_entity`, `max_w` and
+              `bms_soc_floor` (SAFETY-32) are required of every entry - the schema enforces it on the
               Supervisor path, `validate.py` rule 9 enforces it for sim/inject.py
               and the checks, which reach this function with any dict at all.
 
@@ -380,6 +405,12 @@ def build_roster(opts: dict, store, sign_cfg=None) -> list:
             batt_entity=opts.get("batt_entity", ""),
             batt_invert=bool(opts.get("batt_invert")),
             max_w=float(opts.get("max_w", 2000)),
+            # SAFETY-32. THE FLAT KEY, because this IS the single-inverter shape
+            # - the one pack the flat options describe is the one pack whose
+            # cut-out `bms_soc_floor` states. Defaulted from CFG-03's single
+            # mirror exactly as validate.py rule 8 defaults it, so an install
+            # that never set the key reads the same 10 % in both places.
+            bms_soc_floor=_flat_bms_floor(opts),
             store=store, sign_cfg=sign_cfg)]
 
     roster = []
@@ -401,10 +432,54 @@ def build_roster(opts: dict, store, sign_cfg=None) -> list:
             # ⚠️ NOT `opts["max_w"]`. See the docstring: a rating an entry did
             # not state is not the flat one, it is a rating rule 9 refuses.
             max_w=_entry_max_w(entry),
+            # ⚠️ NOT `opts["bms_soc_floor"]`, FOR `max_w`'S REASON AND THIS
+            # CARD'S. An entry that omits its cut-out has NOT got the flat one;
+            # it has none, and rule 8 refuses it by name. Borrowing here would be
+            # SAFETY-32's own defect written in three lines - the shallower pack
+            # graded against the deeper pack's BMS.
+            bms_soc_floor=_entry_bms_floor(entry),
             store=store, sign_cfg=sign_cfg))
     return roster
 
 
+def _flat_bms_floor(opts: dict):
+    """The single-inverter shape's own cut-out, or None if it is unparseable.
+
+    Defaulted from `PEAK_OPTION_DEFAULTS` when the key is absent or blank, which
+    is `validate._number()`'s behaviour on the same key - the two have to agree
+    or the validator would grade a number the roster does not build.
+    """
+    # ⚠️ ONE defaulting line, not two. The first cut wrote the fallback into the
+    # `.get()` AND into the blank test below it, and a mutation that deleted
+    # either one survived - each was covering for the other, which is
+    # MUTATION.md's masking section in three lines of configuration reading.
+    raw = opts.get("bms_soc_floor")
+    if raw is None or raw == "":
+        return float(PEAK_OPTION_DEFAULTS["bms_soc_floor"])
+    try:
+        return float(raw)
+    except (TypeError, ValueError):
+        return None
+
+
+def _entry_bms_floor(entry: dict):
+    """One entry's OWN cut-out, or None when it did not state a usable one.
+
+    None rather than a borrowed number and rather than 0.0 - `_entry_max_w()`
+    can answer 0.0 because a zero rating commands nothing, but a zero FLOOR is
+    the most permissive value the key has and would read as "this pack may be
+    emptied". Rule 8 refuses the missing key by name; until it does, None
+    withholds.
+    """
+    raw = entry.get("bms_soc_floor")
+    if raw is None or raw == "":
+        return None
+    try:
+        return float(raw)
+    except (TypeError, ValueError):
+        return None
+
+
 def _entry_max_w(entry: dict) -> float:
     """One entry's supervised cap, or 0.0 when it did not state a usable one.
 

goodwe_controller/app/maintenance.py

@@ -18,16 +18,47 @@ OWNERSHIP, not priority. Exactly one thing writes the setpoint at any moment:
 The loop is never "disabled" - it yields. The heartbeat to the inverter keeps
 running in every phase, so the hardware watchdog stays fed and a crash here
 still ends with the inverter at 0 W rather than latched.
+
+ONE PACK AT A TIME (MAINT-05). There is now one `Maintenance` per roster entry,
+each with its own schedule, its own phase and its own history under its own
+Store keys - and therefore, for the first time, two of them able to move power
+on ONE grid connection at once. Read `arbitrate()` before touching any of it:
+the site holds exactly ONE maintenance slot, a pack that is not IDLE holds it,
+and every other pack is held at IDLE under the loop until it is released. That
+is not a scheduling nicety. A drain of pack A against a charge of pack B is two
+power-moving phases pulling opposite ways across the meter - the circulating
+power SAFETY-31 exists to DETECT, arriving here as something this module can
+simply not do.
 """
 
 import logging
+import math
 from dataclasses import dataclass, field
 from datetime import datetime, timedelta, timezone
 
+from .store import MAINT_UNIT_KEYS, unit_key
+
 _LOG = logging.getLogger("goodwe.maint")
 
 IDLE, DRAIN, CHARGE, HOLD = "idle", "drain", "charge", "hold"
 
+# ⚠️ THE PHASES THAT MOVE POWER, WHICH IS NOT THE SAME LIST AS "owns the
+# setpoint". The table above is about WHO WRITES the number; this is about
+# whether the pack is being pushed. DRAIN and HOLD own the setpoint. CHARGE does
+# NOT - the loop owns it - and CHARGE is in here anyway, because the charge-only
+# clamp plus the cheap-window floor is what pulls a second pack full while the
+# first is being emptied. Leaving CHARGE out is precisely the interlock hole
+# this card is about: pack A charging from the grid at 2000 W while pack B
+# drains 2000 W to export is 4 kW going round in a circle with the meter at
+# zero, and both packs believing they are behaving.
+#
+# HOLD commands 0 W and moves nothing, and it is in here for the other half of
+# the sentence: it OWNS the actuator - the pack is parked at the top for the BMS
+# and is deliberately not available to the loop - so a second cycle starting
+# under it would still be two recalibrations at once, which is the thing whose
+# value is a clean uninterrupted coulomb count.
+POWER_MOVING = (DRAIN, CHARGE, HOLD)
+
 # A drain of a 13.8 kWh usable pack (300 Ah x 51.2 V, DoD 90) at the shipped
 # 2000 W is ~6.9 h even with no house load helping - and less in practice, since
 # house load drains alongside the export and soc_floor stops short of empty.
@@ -153,17 +184,102 @@ class MaintResult:
     setpoint_w: float | None = None
     charge_only: bool = False
     events: list = field(default_factory=list)
+    # ⚠️ WHY A DEFERRAL IS A FIELD AND NOT JUST A LOG LINE. A pack that is due
+    # and is not starting looks, from every surface an operator has, exactly
+    # like a pack whose schedule has quietly stopped firing - §12's opening
+    # sentence, the failure this whole module exists because of. It is carried
+    # so the status page can SAY "waiting for the other pack" rather than
+    # showing an idle machine and a due date in the past. Empty on every path
+    # that is not deferring, so a roster of one never sees it.
+    deferred: str = ""
+
+
+class UnitHistory:
+    """One pack's maintenance history in the Store, under one spelling.
+
+    ⚠️ THE NAMESPACE IS A CONSTRUCTOR ARGUMENT AND `None` IS A REAL VALUE, NOT
+    AN OMISSION. `unit=None` means the FLAT keys - `phase`, `phase_started`,
+    `last_completed`, `last_start_attempt` - which is what every install has
+    written since the add-on shipped and what a one-inverter site keeps writing
+    after MAINT-05. §6.2's identity: with one pack there is nothing for a
+    namespace to distinguish it FROM, and adding one would change the layout of
+    a live state file to say something it already said. `arbiter.actuator_for()`
+    makes the same move for the same reason and its docstring argues it.
+
+    ⚠️ `mirror_flat` IS THE ROLLBACK VIEW AND IT IS ONLY EVER THE FIRST PACK'S.
+    At N>=2 pack A's history has been migrated into its namespace and is written
+    there; the flat keys are ALSO kept in step, so an operator who rolls the
+    add-on back to a single-FSM version finds a CURRENT history rather than one
+    dated the day the second inverter was declared. store.migrate_maint_keys()
+    carries the argument, and it is DEV-02's rule 2 applied to history instead
+    of to a latch: "left in the file" is not the same promise as "still true".
+    Pack B has no flat spelling to mirror to - the flat keys mean pack A on
+    every install that has them - so it is constructed with mirroring off, and
+    a mirror from B would OVERWRITE A's rollback view with B's phase.
+    """
+
+    def __init__(self, store, unit: str | None = None, mirror_flat: bool = False):
+        self.store = store
+        self.unit = unit
+        self.mirror_flat = bool(mirror_flat)
+
+    def key(self, name: str) -> str:
+        """The Store key this pack's `name` lives under."""
+        return name if self.unit is None else unit_key(self.unit, name)
+
+    def get(self, name: str, default=None):
+        return getattr(self.store, "data", {}).get(self.key(name), default)
+
+    def set(self, name: str, value) -> None:
+        self.store.set(self.key(name), value)
+        if self.mirror_flat:
+            self.store.set(name, value)
+
+    def get_time(self, name: str):
+        return self.store.get_time(self.key(name))
+
+    def set_time(self, name: str, when) -> None:
+        self.store.set_time(self.key(name), when)
+        if self.mirror_flat:
+            self.store.set_time(name, when)
 
 
 class Maintenance:
-    def __init__(self, cfg: MaintConfig, store):
+    """ONE pack's recalibration cycle: its schedule, its phase, its history.
+
+    ⚠️ ONE PER ROSTER ENTRY SINCE MAINT-05, AND THE SLOT IS NOT ITS OWN. This
+    object decides WHEN ITS PACK would like to start; `arbitrate()` decides
+    which of them may, because the constraint is a property of the grid
+    connection and no single FSM can see it. Never call `tick()` in a loop
+    without passing the `blocked_by` that `arbitrate()` returned for it: that is
+    the two-packs-at-once bug with the interlock present but not wired.
+    """
+
+    def __init__(self, cfg: MaintConfig, store, unit: str | None = None,
+                 name: str = "", mirror_flat: bool = False):
         self.cfg = cfg
-        self.store = store
+        # ⚠️ `self.store` IS THE VIEW, NOT THE Store. Everything in this class
+        # names an UNQUALIFIED key - "phase", "last_completed" - and the view is
+        # the one place that turns it into the key this pack owns. A raw
+        # `store.set("phase", ...)` anywhere below would write pack A's history
+        # from pack B's state machine, which is the shared-state bug this card
+        # is about, and there would be nothing to see at a roster of one.
+        self.store = UnitHistory(store, unit, mirror_flat)
+        self.unit = unit
+        # What an event or a refusal calls this pack. At a roster of one it is
+        # never rendered - a single machine cannot be deferred - so the shipped
+        # one-inverter reason strings are untouched.
+        self.name = str(name or unit or "the pack")
+        # Edge-memo for the deferral note. The action log holds 40 lines and a
+        # deferral lasts as long as the other pack's cycle - hours - so a line
+        # per cycle would flush the very context that explains it. Same pattern
+        # as `Controller._arb_action` and `_maint_shape_suspended`.
+        self._deferred = ""
 
     # -- state ---------------------------------------------------------------
     @property
     def phase(self) -> str:
-        return self.store.data.get("phase", IDLE)
+        return self.store.get("phase", IDLE)
 
     def _enter(self, phase: str, now: datetime, events: list, msg: str) -> None:
         self.store.set("phase", phase)
@@ -185,14 +301,60 @@ class Maintenance:
         last = self.store.get_time("last_completed")
         return None if last is None else last + timedelta(days=self.cfg.interval_days)
 
+    def waited_days(self, now: datetime) -> float:
+        """Days since this pack last COMPLETED a cycle; `inf` if it never has.
+
+        ⚠️ THE ARBITRATION KEY, AND IT IS DELIBERATELY NOT `due()`. `due()` is a
+        yes/no and two packs that are both due are indistinguishable by it -
+        which is how a roster-order tie-break gets written and how the first
+        entry then wins every contested month forever while the second drifts.
+        This is an ORDERING: the pack whose coulomb counter has gone longest
+        without a recalibration goes first. `inf` for never-recalibrated is the
+        same statement `due()` makes about an absent `last_completed`, kept as
+        one so a fresh pack cannot lose to one that ran last month.
+        """
+        last = self.store.get_time("last_completed")
+        return math.inf if last is None else (now - last).total_seconds() / 86400.0
+
+    def _cooled(self, now: datetime) -> bool:
+        """Is the post-abort cooldown over?"""
+        last_try = self.store.get_time("last_start_attempt")
+        return (last_try is None
+                or (now - last_try) >= timedelta(minutes=RESTART_COOLDOWN_MIN))
+
+    def wants_start(self, now: datetime) -> bool:
+        """Would this pack begin a cycle on this tick if the site let it?
+
+        ⚠️ THE SAME PREDICATE `tick()` USES, NOT A COPY OF IT. `arbitrate()` has
+        to know who is asking before anybody moves, and a second transcription
+        of "enabled, in the start hour, due and cooled down" is a second opinion
+        about who was asking - so the IDLE branch of `tick()` calls this one.
+        """
+        return (self.phase == IDLE and self.cfg.enabled
+                and now.hour == self.cfg.start_hour and self.due(now)
+                and self._cooled(now))
+
     # -- operations ----------------------------------------------------------
-    def force_start(self, now: datetime) -> list:
+    def force_start(self, now: datetime, blocked_by: str = "") -> list:
         """Start a cycle regardless of the calendar.
 
         Commissioning uses this: a cycle must be proven end to end before a site
         is signed off, and nobody waits a month to find out the schedule never
         fires.
+
+        ⚠️ REGARDLESS OF THE CALENDAR, NOT REGARDLESS OF THE INTERLOCK. The
+        calendar is a preference and an operator may overrule it; "one pack
+        moves power at a time" is a property of the grid connection and an
+        operator pressing a button does not add a second one. A forced start
+        against a busy site is REFUSED, by name, and the phase machine does not
+        move - the refusal is returned as an event so the operator reads it in
+        the action log rather than watching nothing happen.
         """
+        if blocked_by:
+            msg = (f"maintenance: forced start REFUSED on {self.name} - "
+                   f"{blocked_by}")
+            _LOG.warning("%s", msg)
+            return [msg]
         events = []
         self.store.set_time("last_start_attempt", now)
         self._enter(DRAIN, now, events, "maintenance: forced start -> drain")
@@ -203,23 +365,80 @@ class Maintenance:
         self._enter(IDLE, now, events, f"maintenance: ABORT ({why})")
         return events
 
+    def _defer_note(self, why: str, events: list) -> list:
+        """Say once, not every second, that this pack is waiting its turn.
+
+        ⚠️ THE WORDING IS DELIBERATELY NOT AN ABORT AND NOT A TIMEOUT. §12's
+        notes are explicit that a 12 h phase timeout means "the battery is not
+        following commands - a FAULT, not a slow day", and the one thing that
+        must not happen here is an operator reading a deferral as that. This
+        line says the pack is WAITING, names the pack it is waiting for, and
+        says it will run next - none of which any failure path in this module
+        says.
+        """
+        if self._deferred != why:
+            self._deferred = why
+            msg = (f"maintenance: {self.name} is due and is WAITING its turn - "
+                   f"{why}. It is idle under the control loop and will "
+                   f"recalibrate at the next start hour once the slot is free; "
+                   f"this is not a stalled phase.")
+            _LOG.info("%s", msg)
+            events.append(msg)
+        return events
+
     # -- the machine ---------------------------------------------------------
-    def tick(self, now: datetime, soc: float | None) -> MaintResult:
+    def tick(self, now: datetime, soc: float | None,
+             blocked_by: str = "") -> MaintResult:
+        """One cycle of this pack's state machine.
+
+        `blocked_by` is `arbitrate()`'s answer for THIS pack: empty when the
+        site's one maintenance slot is available to it, otherwise the reason it
+        is not. It gates two things and nothing else - a start, and a phase that
+        is already moving power on a pack that does not hold the slot.
+        """
         events: list = []
         phase = self.phase
 
+        # ⚠️ THE SLOT IS CHECKED BEFORE THE PHASE, AND THIS BRANCH IS A RECOVERY
+        # RATHER THAN AN EXPECTED PATH. `arbitrate()` grants the slot to the
+        # pack that is already mid-cycle, so a pack that is BOTH moving power
+        # and blocked means two histories on disk say "drain" - a hand-edited
+        # state file, or a store carried across a version that had no interlock.
+        # The fail-safe direction is unambiguous: two packs pulling against each
+        # other on one connection is the fault SAFETY-31 latches the whole site
+        # for. Stand this one down to 0 W and say why; the holder keeps its
+        # cycle, which is the outcome that loses no coulomb count at all.
+        if blocked_by and phase in POWER_MOVING:
+            events += self.abort(now, f"interlock - {blocked_by}")
+            return MaintResult(IDLE, True, 0.0, False, events)
+
         if phase == IDLE:
-            if self.cfg.enabled and now.hour == self.cfg.start_hour and self.due(now):
-                last_try = self.store.get_time("last_start_attempt")
-                cooled = (
-                    last_try is None
-                    or (now - last_try) >= timedelta(minutes=RESTART_COOLDOWN_MIN)
-                )
-                if cooled:
-                    self.store.set_time("last_start_attempt", now)
-                    self._enter(DRAIN, now, events,
-                                "maintenance: due -> drain (exports the surplus)")
-                    return MaintResult(DRAIN, True, self.cfg.discharge_w, False, events)
+            if self.wants_start(now):
+                if blocked_by:
+                    # ⚠️ `last_start_attempt` IS NOT STAMPED HERE, AND THAT IS
+                    # THE STARVATION FIX RATHER THAN AN OMISSION. That key is
+                    # the post-ABORT cooldown: it exists so a cycle that failed
+                    # does not re-trigger inside the same start hour. A deferral
+                    # is not an attempt - nothing was tried and nothing failed -
+                    # and stamping it would put this pack under a 90 minute
+                    # cooldown for a start hour that is 60 minutes long, so the
+                    # pack that lost the arbitration would still be locked out
+                    # at the moment the winner released the slot.
+                    #
+                    # ⚠️ AND THE PHASE DOES NOT MOVE, WHICH IS THE OTHER HALF.
+                    # `phase_started` is untouched, so the 12 h DRAIN and 36 h
+                    # CHARGE timeouts - which mean "the battery is not following
+                    # commands, a fault" - cannot be reached by waiting. A
+                    # deferred pack is IDLE under the loop, not a stuck one.
+                    return MaintResult(IDLE, False, None, False,
+                                       self._defer_note(blocked_by, events),
+                                       blocked_by)
+                self._deferred = ""
+                self.store.set_time("last_start_attempt", now)
+                self._enter(DRAIN, now, events,
+                            "maintenance: due -> drain (exports the surplus)")
+                return MaintResult(DRAIN, True, self.cfg.discharge_w, False, events)
+            self._deferred = ""
             return MaintResult(IDLE, False, None, False, events)
 
         # ⚠️ Fail toward inaction. Without a state-of-charge reading none of the
@@ -273,3 +492,77 @@ class Maintenance:
         _LOG.error("maintenance: unknown phase %r, forcing idle", phase)
         events += self.abort(now, f"unknown phase {phase!r}")
         return MaintResult(IDLE, True, 0.0, False, events)
+
+
+def _days(n: float) -> str:
+    return "never recalibrated" if n == math.inf else f"{n:.0f} d since its last cycle"
+
+
+def arbitrate(machines, now: datetime) -> list:
+    """Grant the site's ONE maintenance slot. Returns a list parallel to
+    `machines`: "" for the pack that may move power, otherwise the reason the
+    pack may not.
+
+    ⚠️ THE SLOT IS A PROPERTY OF THE GRID CONNECTION, WHICH IS WHY IT IS NOT ON
+    `Maintenance`. Each FSM can see its own pack's schedule and nothing else; the
+    quantity being rationed - "how many phases are moving power across one
+    meter" - is a sum over the roster, exactly like `sum(|B_i|) - |sum(B_i)|` in
+    SAFETY-31 and for the same reason. A per-pack answer to a site-level
+    question is the shape of defect this repository keeps finding.
+
+    ⚠️ IT IS DERIVED, NOT PERSISTED, AND THAT IS DELIBERATE. The holder is
+    "whichever pack's stored phase is not IDLE" - a fact already on disk, in the
+    per-pack history the packs themselves write. A separate `maint_slot_holder`
+    key would be a second opinion about the same thing, and the two would
+    disagree the first time the add-on was killed between writing the phase and
+    writing the holder. Restart-safe by construction: a controller that comes
+    back up mid-drain reads the drain and re-derives the holder from it.
+
+    TWO CLAUSES, IN ORDER:
+
+    1. A PACK ALREADY MOVING POWER KEEPS THE SLOT. Nothing preempts a cycle in
+       flight, because a recalibration interrupted halfway is worth less than
+       nothing - the coulomb counter is graded on ONE uninterrupted drain and
+       refill, so a cycle stopped at 40 % has spent the energy and bought no
+       calibration. If more than one pack somehow comes up moving power, the
+       FIRST in roster order holds and `tick()` stands the others down; see the
+       recovery branch there.
+
+    2. OTHERWISE THE LONGEST-WAITING CANDIDATE STARTS. `wants_start()` is asked
+       of every pack, and among those that say yes the slot goes to the one with
+       the largest `waited_days()`. NOT roster order: two packs on one monthly
+       schedule are due in the same hour every month, and a roster-order tie
+       would hand the slot to `inverters[0]` every single time while the second
+       pack's coulomb counter drifted forever - which is, in a different
+       costume, the exact defect this card was opened for. Roster order breaks a
+       genuine numeric tie and nothing else.
+
+    ⚠️ AT A ROSTER OF ONE THIS IS THE IDENTITY FUNCTION. One machine is never
+    another machine's blocker: clause 1 finds at most itself and clause 2 has at
+    most one candidate, so it returns `[""]` on every tick and no reason string
+    it can build is ever rendered. That is what makes the one-inverter behaviour
+    byte-identical rather than merely similar.
+    """
+    blocks = [""] * len(machines)
+    busy = [(i, m) for i, m in enumerate(machines) if m.phase in POWER_MOVING]
+    if busy:
+        held, holder = busy[0]
+        for i, m in enumerate(machines):
+            if i != held:
+                blocks[i] = (f"{holder.name} is mid-cycle ({holder.phase}) and "
+                             f"holds the site's one maintenance slot")
+        return blocks
+
+    wants = [(i, m) for i, m in enumerate(machines) if m.wants_start(now)]
+    if len(wants) <= 1:
+        return blocks
+    # Largest waited_days wins; `-index` makes a numeric tie fall to roster
+    # order rather than to whatever `max()` happens to see first.
+    won, winner = max(wants, key=lambda p: (p[1].waited_days(now), -p[0]))
+    for i, m in wants:
+        if i != won:
+            blocks[i] = (f"{winner.name} has waited longer "
+                         f"({_days(winner.waited_days(now))} against "
+                         f"{_days(m.waited_days(now))}) and takes the site's "
+                         f"one maintenance slot first")
+    return blocks

goodwe_controller/app/peakguard.py

@@ -360,6 +360,29 @@ distinction in DOCS.md and in this comment. Revisit if a THIRD floor appears -
 a loop-side guard would be the third - at which point the migration buys enough
 to be worth its cost.
 
+------------------------------------------------------------------------------
+SAFETY-32: ONE `bms_soc_floor` FOR TWO PACKS GRADES ONE AGAINST THE OTHER'S BMS
+
+`bms_soc_floor` above describes A PACK, not a site, and it was a flat option
+because there was one pack. At two, two batteries of different models have
+different cut-outs, and one number for both is wrong in the direction that
+empties a battery: the SHALLOWER pack (the higher cut-out) is protected by a
+floor that belongs to the DEEPER one. So it became a roster key - `Pack` carries
+each unit's own value, `pack_floor_pct()` conjoins it with the site's EMS floor,
+and validate.py rule 8 grades EVERY unit and NAMES the one it refuses.
+
+⚠️ `peak_soc_floor` DID NOT MOVE, AND THE RESIDUAL IS RECORDED RATHER THAN
+HIDDEN. It is one number for the site, so with cut-outs of 10 % and 18 % the
+operator must set it above 18 and the deeper pack is then withheld above its own
+need. That half - a per-pack EMS floor - is NOT this card and must not be
+smuggled in with it; DECISIONS/SAFETY-32-per-pack-bms-floor.md states it as the
+open question it is.
+
+⚠️ AND THE MIN() OVER MEASURED SoC IS STILL RIGHT. `Controller.pack_floor_soc`
+reduces to the emptiest pack and this card does not touch it. What changed is
+the FLOOR it is graded against, and the addition of a per-pack answer beside the
+site's one verdict - see `pack_permissions()`.
+
 ------------------------------------------------------------------------------
 AND IT READS NO PER-PHASE QUANTITY, ON PURPOSE
 
@@ -572,6 +595,110 @@ def discharge_permitted(soc, soc_floor) -> str | None:
     return None
 
 
+@dataclass(frozen=True)
+class Pack:
+    """SAFETY-32. ONE pack's three numbers for the floor question, NAMED.
+
+    `name`       the roster name, because a refusal that does not say WHICH pack
+                 stopped the site sends an operator to the wrong device - the
+                 same defect CFG-04's `unit_caps_indexed()` was written for.
+    `soc`        THIS pack's measured state of charge. None when its sensor did
+                 not answer, and None is not a full battery.
+    `bms_floor`  THIS pack's OWN cut-out percent (`bms_soc_floor`), which is a
+                 property of the hardware in the cupboard and NOT of the site.
+                 None when the roster entry did not state one - see
+                 `pack_floor_pct()` for why that is withheld and not 0 %.
+
+    ⚠️ IT CARRIES THE CUT-OUT, NOT THE EMS FLOOR. `peak_soc_floor` is one number
+    for the whole site (the depth a money-saving discharge may never reach) and
+    stays a flat option; `bms_soc_floor` describes a battery and became a roster
+    key here, because two packs of different models have different cut-outs and
+    grading both against one of them is wrong in the direction that empties the
+    shallower one.
+    """
+
+    name: str
+    soc: float | None
+    bms_floor: float | None
+
+
+def pack_floor_pct(pack, soc_floor):
+    """The percent THIS pack may not be spent below, or None if that is unknown.
+
+    `max(the site's EMS floor, this pack's own cut-out)` - the CONJUNCTION, and
+    the direction matters: the EMS floor is what the operator asked for and the
+    cut-out is what the hardware will do anyway, so a pack is protected by
+    whichever is higher. Taking the site floor alone is the shipped defect this
+    card removes (the shallower pack graded against the deeper one's cut-out);
+    taking the cut-out alone would throw SAFETY-24's margin away.
+
+    ⚠️ AN UNSTATED CUT-OUT IS None, NOT ZERO, AND NOT THE FLAT OPTION. 0 % would
+    be a floor that permits everything - the most permissive value there is,
+    invented from an absence - and borrowing the flat `bms_soc_floor` is CFG-04's
+    "nothing is inherited" broken in one line, which is this card's whole subject
+    restated. `discharge_permitted()` already treats a None floor as NO
+    permission, so an entry that forgot the key costs a shave rather than a pack.
+
+    ⚠️ AND UNDER A rule-8-VALID CONFIG THIS IS `soc_floor`, FOR EVERY PACK.
+    validate.py rule 8 refuses `peak_soc_floor <= bms_soc_floor` on EVERY unit,
+    so `max()` selects the EMS floor everywhere the add-on will actually start -
+    which is exactly why the reduction is written here rather than assumed at the
+    call site. "Another rule refuses that config" is a claim about today's rule
+    ordering, not a property of this function, and `validate_startup_config()` is
+    also `sim/inject.py`'s entry point. `unit_caps_indexed()`'s docstring records
+    the same argument, and CFG-04's index desynchronisation is what happened the
+    last time it was left implicit.
+    """
+    if soc_floor is None or pack.bms_floor is None:
+        return None
+    return max(float(soc_floor), float(pack.bms_floor))
+
+
+def pack_permissions(packs, soc_floor) -> list[tuple[str, str | None]]:
+    """SAFETY-32. Per pack: None if it may be spent, else the reason it may not.
+
+    ONE `(name, reason)` pair per pack, in roster order, from the SAME
+    `discharge_permitted()` the site verdict uses - a second transcription of
+    "may this pack be spent" is a second answer to it, which is the argument
+    that made that predicate public in the first place.
+
+    ⚠️ THIS IS THE OBSERVABLE THE MIN() CANNOT PRODUCE. `Controller.pack_floor_soc`
+    reduces the measured SoC to the emptiest pack BEFORE anything is graded, and
+    that reduction is right for the site's one verdict - but it has thrown away
+    which pack it came from and how far that pack was from ITS OWN floor by the
+    time the answer exists. Two packs at 15 % against floors of 10 % and 20 %
+    have opposite answers and are indistinguishable after the `min()`.
+    """
+    return [(p.name, discharge_permitted(p.soc, pack_floor_pct(p, soc_floor)))
+            for p in packs]
+
+
+def binding_soc_floor(packs, soc_floor):
+    """The ONE floor `decide()` may grade `pack_floor_soc` against, at a roster.
+
+    MAXIMUM over the packs' own floors, and it is the conservative reduction that
+    pairs with the MINIMUM `pack_floor_soc` reduction: `min(soc) <= max(floor)`
+    is implied by "some pack is at or below its own floor", so this verdict is
+    never more permissive than grading each pack separately would be. The other
+    direction is not true and is not claimed - a site can be refused here while
+    every pack individually has headroom, which costs a shave and never a pack.
+
+    None when ANY pack's floor is unknown, for `Controller.site_dev_bound()`'s
+    reason and in the same direction: `discharge_permitted()` treats a None floor
+    as no permission, so an unknown floor withholds rather than defaults.
+
+    ⚠️ AT N=1 THIS IS `soc_floor` ITSELF under any config rule 8 accepts, and
+    test_peakguard.py part 12 asserts that over the whole schema range rather
+    than assuming it.
+    """
+    if not packs:
+        return None if soc_floor is None else float(soc_floor)
+    floors = [pack_floor_pct(p, soc_floor) for p in packs]
+    if any(f is None for f in floors):
+        return None
+    return max(floors)
+
+
 def decide(projection, *, batt_w, cap_w, reading, achievable_w, soc, soc_floor,
            max_hold_s=None, frozen: bool = False) -> PeakDecision:
     """The whole guard. Pure: no clock, no I/O, no state carried between calls.

goodwe_controller/app/store.py

@@ -71,6 +71,24 @@ _LOG = logging.getLogger("goodwe.store")
 # the card is gone while the fault runs on - exactly how the 2026-09-02 outage
 # went unseen for 6 h 25 m with the detector correct throughout. main.py
 # re-raises the notification from this key on every start.
+#
+# ⚠️ AND SAFETY-30's PAIR IS FLAT FOR THE SAME REASON SAFETY-31's IS, WHICH IS
+# NOT THE SAME REASON THE ONES ABOVE ARE NAMESPACED. `site_sign_tripped` /
+# `site_sign_reason` hold the SITE sign-disagreement latch: the SUM of every
+# standing command opposed the meter, continuously. That is a statement about
+# the grid connection - the offending number is the site target, which no single
+# inverter owns - so there is no unit to hang it on, and hanging it on one would
+# let an operator clear "inverter B's half" of a fault that has no halves.
+# `unit/<name>/sign_tripped` above stays what it always was and now means what
+# it says: THAT unit's own command against THAT unit's own measured power.
+#
+# ⚠️ AND THEY ARE NOT IN `LEGACY_UNIT_KEYS`, MUST NEVER BE, AND ARE NOT WRITTEN
+# AT A ROSTER OF ONE AT ALL. At N=1 the site comparison and the unit comparison
+# are the same sentence and `Controller.cycle()` arms only the per-unit
+# detector, so a one-inverter install's state file is byte-identical to what it
+# was before this card - these two keys sit at their defaults and nothing reads
+# them. A rollback therefore finds exactly the flat `sign_tripped` mirror it
+# always found.
 DEFAULTS = {
     "phase": "idle",
     "phase_started": None,
@@ -80,6 +98,8 @@ DEFAULTS = {
     "sign_reason": "",
     "circ_tripped": False,
     "circ_reason": "",
+    "site_sign_tripped": False,
+    "site_sign_reason": "",
 }
 
 # --------------------------------------------------------------------------- #
@@ -106,16 +126,53 @@ SCHEMA_VERSION = 1
 SCHEMA_KEY = "schema_version"
 
 # Which flat keys belong to ONE inverter, and therefore move into its namespace.
-# ⚠️ THE MAINTENANCE KEYS ARE NOT IN HERE AND THAT IS DELIBERATE. `phase`,
-# `phase_started`, `last_completed` and `last_start_attempt` belong to the
-# maintenance state machine, of which this add-on still holds exactly ONE - a
-# monthly recalibration is a pack procedure and per-pack SCHEDULING is a card of
-# its own, not something to infer from a key rename. Moving them here would
-# quietly re-home an install's whole maintenance history under a unit name; they
-# stay where they are, untouched, which is the other half of "migrates without
-# losing its latch OR ITS HISTORY".
+# ⚠️ THE MAINTENANCE KEYS ARE STILL NOT IN HERE, AND SINCE MAINT-05 THAT IS A
+# DIFFERENT DECISION RATHER THAN THE SAME ONE. They DO move now - one
+# `Maintenance` per roster entry needs one history per roster entry - but they
+# move on their OWN marker, `MAINT_SCHEMA_KEY`, and only on a site that actually
+# has more than one pack. Two conditions the SAFETY-02 keys do not share, so
+# folding them into this tuple would either drag them across on a
+# single-inverter install (which is the byte-identical store layout MAINT-05
+# must not change) or hold the sign latch hostage to a roster count it has
+# nothing to do with. `MAINT_UNIT_KEYS` below carries them.
 LEGACY_UNIT_KEYS = ("sign_tripped", "sign_reason")
 
+# --------------------------------------------------------------------------- #
+# MAINT-05: THE SECOND MIGRATION, AND WHY IT IS A SECOND ONE
+#
+# 0 = the flat, one-history layout every install has had since the add-on
+#     shipped: `phase` / `phase_started` / `last_completed` /
+#     `last_start_attempt` describe THE pack, because there was one FSM.
+# 1 = one history per roster entry, under `unit/<name>/...`.
+#
+# ⚠️ ITS OWN MARKER, NOT A BUMP OF `SCHEMA_VERSION`, AND THE REASON IS THAT THE
+# TWO MIGRATIONS DO NOT HAVE THE SAME PRECONDITION. DEV-02's runs on every
+# install exactly once and is monotone in the version number. This one runs ONLY
+# WHERE THE ROSTER HAS MORE THAN ONE PACK, because at N=1 §6.2's identity holds
+# and the flat keys ARE that pack's namespace - moving them would add four keys
+# and a version marker to a store MAINT-05's own acceptance criterion requires
+# to be byte-identical. A single monotone counter cannot express "migrated on
+# this site, deliberately not migrated on that one" without lying to the next
+# version that reads it: a one-inverter store would have to be left at the old
+# number forever, and then a genuine v2 migration could never be told apart from
+# a site that simply never grew.
+#
+# ⚠️ AND IT IS TWO MARKERS, NOT TWO COPIES OF THE THREE RULES.
+# `_move_flat_keys()` and `_verify_moves()` below are shared verbatim by both
+# migrations, so rules 1 and 3 and the READ-BACK POSTCONDITION are one
+# implementation. A second transcription of a migration is a second opinion
+# about what "it landed" means.
+MAINT_SCHEMA_VERSION = 1
+MAINT_SCHEMA_KEY = "maint_schema_version"
+
+# The flat maintenance-history keys, in the order they are moved.
+# ⚠️ THE VALUES ARE THE FLAT SPELLINGS AND THE MODULE THAT READS THEM INDEXES
+# THIS SAME TUPLE (`maintenance.UnitHistory`), so a key renamed here cannot
+# silently stop being migrated - the same property `UNIT_STORE_KEYS` gives the
+# SAFETY-02 pair.
+MAINT_UNIT_KEYS = ("phase", "phase_started", "last_completed",
+                   "last_start_attempt")
+
 
 def unit_key(unit: str, key: str) -> str:
     """The namespaced Store key for `key` on inverter `unit`.
@@ -154,16 +211,35 @@ class MigrationReport:
         # to shout: an alarm that fires on the harmless case is an alarm an
         # operator learns to dismiss before the real one arrives.
         self.lost = []
+        # ⚠️ A DIFFERENT FAILURE FROM `lost`, AND IT HAS TO BE, BECAUSE THE
+        # READ-BACK CANNOT SEE IT. `lost` is "the store took the write and
+        # dropped it". This is "the store took the write, kept it perfectly, and
+        # NOTHING WILL EVER READ IT" - the migration was handed a unit name that
+        # is not on the roster, so every machine looks under a different key and
+        # finds nothing. There is no harmless-default version of this the way
+        # there is for `lost`: a write nobody reads is not information, whatever
+        # its value, so anything in here fails `ok` on its own.
+        self.misdirected = []
+        # Why nothing ran, when nothing ran. Empty means "already at schema".
+        self.skipped = ""
 
     @property
     def ok(self) -> bool:
-        """Nothing that CARRIED STATE failed to land."""
-        return not self.lost
+        """Nothing that CARRIED STATE failed to land, and it landed WHERE THE
+        MACHINES LOOK."""
+        return not self.lost and not self.misdirected
 
     def explain(self) -> str:
         if not self.ran:
-            return (f"store already at schema {self.from_version}; no "
-                    f"per-inverter migration needed")
+            # ⚠️ `skipped` RATHER THAN A SECOND SENTENCE ABOUT VERSIONS. Since
+            # MAINT-05 there are two reasons a migration does not run - the
+            # store is already at this schema, or the roster has one pack and
+            # the flat keys ARE its namespace - and an operator reading the
+            # startup log needs to be told which, because only one of them
+            # changes when a second inverter is declared.
+            return self.skipped or (f"store already at schema "
+                                    f"{self.from_version}; no "
+                                    f"per-inverter migration needed")
         out = (f"migrated {len(self.moved)} flat key(s) to inverter "
                f"namespaces: " + ", ".join(sorted(self.moved)))
         if self.kept:
@@ -172,6 +248,8 @@ class MigrationReport:
             out += " - PROBLEMS: " + "; ".join(self.problems)
         if self.lost:
             out += " - STATE LOST: " + "; ".join(self.lost)
+        if self.misdirected:
+            out += " - WRONG NAMESPACE: " + "; ".join(self.misdirected)
         return out
 
 
@@ -223,7 +301,30 @@ def migrate_unit_keys(store, unit: str) -> MigrationReport:
         return rep
 
     rep.ran = True
-    for legacy in LEGACY_UNIT_KEYS:
+    _move_flat_keys(store, unit, LEGACY_UNIT_KEYS, rep)
+    _verify_moves(store, rep)
+
+    if rep.ok:
+        store.set(SCHEMA_KEY, SCHEMA_VERSION)
+    else:
+        # ⚠️ NOT MARKED DONE. A migration that lost state must run again on the
+        # next start rather than be recorded as finished - the one thing worse
+        # than a failed migration is a failed migration a later version skips.
+        _LOG.error("per-inverter store migration LOST STATE and has not been "
+                   "marked done: %s", rep.explain())
+    return rep
+
+
+def _move_flat_keys(store, unit: str, keys, rep: MigrationReport) -> None:
+    """Rule 1, once, for both migrations.
+
+    Copies each flat `key` to `unit/<unit>/<key>` UNLESS the namespaced spelling
+    is already present - a store that has both has been written by a version
+    that knew about the namespace, and that value is the live one. Rule 2 is the
+    absence of a delete: the flat keys are left exactly where they were.
+    """
+    data = getattr(store, "data", {})
+    for legacy in keys:
         target = unit_key(unit, legacy)
         if target in data:
             rep.kept.append(target)
@@ -232,17 +333,63 @@ def migrate_unit_keys(store, unit: str) -> MigrationReport:
         # tripped has no `sign_tripped` in its file at all, and the value it
         # then lands on has to be the one `Store.load()` would have supplied -
         # otherwise "absent" and "never tripped" stop being the same statement,
-        # which is the distinction store.py's own header is about.
+        # which is the distinction store.py's own header is about. The same
+        # sentence holds for `last_completed`: absent means "never recalibrated"
+        # and a literal here would be a second opinion about that.
         value = data.get(legacy, DEFAULTS.get(legacy))
         store.set(target, value)
         rep.moved[target] = value
 
-    # THE POSTCONDITION, AND IT IS A READ-BACK RATHER THAN AN ASSUMPTION. This
-    # is the difference between this function and a two-line copy: the caller
-    # can ask "is the latch that was in the file before now under the key the
-    # watchdog reads", through the same `unit_key()` the watchdog's constructor
-    # reads through, and a `False` there is a guard that is off on an install
-    # whose state file says it is on.
+
+def _verify_moves(store, rep: MigrationReport, roster=None) -> None:
+    """THE POSTCONDITION, AND IT IS A READ-BACK RATHER THAN AN ASSUMPTION.
+
+    This is the difference between these functions and a two-line copy: the
+    caller can ask "is the latch that was in the file before now under the key
+    the watchdog reads" - and, since MAINT-05, "is the history that was in the
+    file before now under the key THIS PACK's state machine reads" - through the
+    same `unit_key()` those readers read through. A `False` here is a guard that
+    is off, or a month of coulomb-count history gone, on an install whose state
+    file says otherwise.
+
+    ⚠️ AND `roster` IS THE OTHER HALF, WHICH THE READ-BACK ON ITS OWN CANNOT BE.
+    A POSTCONDITION THAT READS BACK THE NAMESPACE IT WAS HANDED CANNOT DETECT
+    BEING HANDED THE WRONG ONE. That is not a hypothetical: MAINT-05 shipped
+    with `migrate_maint_keys()` taking a bare `unit` string, and a call site that
+    passed `inverter.UNIT_A` instead of the first roster entry's name migrated an
+    install's whole maintenance history into `unit/inverter-a/...` on a roster of
+    `["garage", "shed"]`, read all four keys back under that same name, was
+    SATISFIED, and stamped the schema marker - so the next start skipped the
+    migration too and both packs came up believing they had never been
+    recalibrated. Every symptom of a clean migration, and the history gone.
+    `build_roster()` names a declared roster's entries from the config and falls
+    back to a POSITIONAL `inverter-N`, so `inverter-a` is not even the default
+    there; there is no roster on which that write could have been read.
+
+    So the namespace is graded against the one the MACHINES read - the roster -
+    rather than against the argument this function was handed, which is the
+    thing under test. A unit that is not on the roster is `misdirected`, which
+    fails `ok`, which means the caller does NOT stamp the version marker and the
+    next start runs the migration again. A re-run is recoverable, by rule 1; a
+    silent skip is the state store.py's header says must never happen.
+
+    ⚠️ AND THE GRADED SET IS `moved` UNION `kept`, WHICH IS WHAT MAKES THAT
+    PARAGRAPH TRUE ON THE SECOND START AS WELL AS THE FIRST. Graded over `moved`
+    alone the refusal lasted exactly one start: the wrong call site is still
+    wrong next time, finds the keys it wrote already present, and rule 1 routes
+    every one of them to `kept` - so `moved` is empty, this loop sees nothing,
+    `ok` comes back True and the marker is stamped on a migration that has
+    reached no readable key. The measurement is in the comment on the loop
+    itself. `test_maintenance.py` section 9c calls this twice with the SAME
+    wrong `unit` over one store, which is the shape nothing in the suite had.
+
+    `roster=None` means "not graded", and the one caller that passes it is
+    `migrate_unit_keys()`, which runs BEFORE `build_roster()` for the reason its
+    own docstring gives - the `Inverter` objects restore their latch in their
+    constructors, so the keys have to be in place before there is a roster to
+    check against. That gap is real and is recorded as a running check rather
+    than left to a comment: see test_maintenance.py section 13.
+    """
     for target, value in rep.moved.items():
         got = getattr(store, "data", {}).get(target, "<<missing>>")
         if got == value:
@@ -254,14 +401,142 @@ def migrate_unit_keys(store, unit: str) -> MigrationReport:
         if value != DEFAULTS.get(legacy):
             rep.lost.append(note)
 
+    if roster is None:
+        return
+    names = [str(n) for n in roster]
+    # `unit_key()` builds `unit/<unit>/<key>`, so the unit is field 1. Split
+    # through the same spelling the writer used rather than re-deriving it.
+    #
+    # ⚠️ `moved` UNION `kept`, AND THE UNION IS THE WHOLE OF THE SECOND FIX
+    # ROUND. Grading only `moved` made the refusal last EXACTLY ONE START.
+    # Measured at Controller level over one persisted store, roster
+    # ["garage", "shed"], flat `last_completed` 2026-08-07, with the call site
+    # still handing `inverter-a` on both starts:
+    #
+    #   construction 1: ok=False misdirected=4 moved=4 kept=0  marker=<absent>
+    #   construction 2: ok=True  misdirected=0 moved=0 kept=4  marker=1
+    #
+    # On start 2 the four keys the WRONG start 1 wrote are already present, so
+    # rule 1 routes every one of them to `rep.kept`, `rep.moved` is EMPTY, this
+    # loop iterates over nothing, `misdirected` is empty and `ok` is True - so
+    # the caller stamps the marker and start 3 takes the `from_version >=
+    # SCHEMA` early return FOREVER, including after the call site is fixed.
+    # Loud-and-recoverable became silent-and-permanent after one restart: the
+    # exact state this module's header says must never happen, reached by the
+    # fix that was written to prevent it.
+    #
+    # The union is sound by construction rather than by hope: every entry in
+    # both `rep.moved` and `rep.kept` is a `unit_key(unit, legacy)` built from
+    # the SAME `unit` argument, so grading `kept` grades the same namespace
+    # decision `moved` does. A correct call site puts its `kept` entries under a
+    # roster name and they pass; a wrong one cannot launder its namespace by
+    # having already written it.
+    for target in sorted(set(rep.moved) | set(rep.kept)):
+        parts = target.split("/")
+        unit = parts[1] if len(parts) > 2 else ""
+        if unit in names:
+            continue
+        rep.verified = False
+        note = (f"{target} was migrated into namespace {unit!r}, which is NOT "
+                f"on the roster {names!r} - nothing reads that key")
+        rep.problems.append(note)
+        rep.misdirected.append(note)
+
+
+def migrate_maint_keys(store, unit: str, roster) -> MigrationReport:
+    """Move the flat maintenance history into the FIRST pack's namespace.
+
+    MAINT-05. Until this card there was one `Maintenance` and one history, and
+    `store.py`'s own header argued the keys should stay flat because per-pack
+    scheduling was a card of its own. This is that card: there is now one FSM
+    per roster entry, so there is one history per roster entry, and the flat one
+    belongs to the pack that was being recalibrated - `inverters[0]`, which is
+    what `Controller.maint_unit` named and what `validate.py` rule 6 graded
+    against.
+
+    ⚠️ `len(roster) <= 1` IS A RETURN, NOT A SPECIAL CASE. At one inverter the
+    flat keys ARE that pack's namespace (§6.2's identity, the same one
+    `arbiter.actuator_for()` leans on), `Maintenance` is constructed to read
+    them directly, and a migration would add four keys plus a marker to a store
+    that must come out byte-identical. The migration happens on the day the
+    roster grows, which is the day the flat spelling stops being able to answer
+    "which pack".
+
+    ⚠️ `roster` IS REQUIRED, AND IT REPLACED AN `n_units: int`. Two reasons, and
+    the second is the one MAINT-05 was rejected over.
+
+    The small one: `n_units` was `len(self.inverters)` at the only call site, so
+    the roster carries it - `len(roster)` - and a count passed beside the names
+    is a second opinion about how many packs there are.
+
+    The load-bearing one: THE NAMES ARE THE POSTCONDITION. `unit` says which
+    pack the flat keys belonged to and that is genuinely a decision - the first
+    roster entry, because that is the pack the one FSM was driving - but a
+    decision the caller can get WRONG, and until this signature existed nothing
+    could tell. `_verify_moves()` read back the namespace it was handed, so a
+    migration into a name no machine reads verified clean and stamped the
+    marker. Handing the roster in means the read-back is graded against the
+    keys the MACHINES read, and a `unit` that is not among these names comes
+    back `misdirected`: not `ok`, marker NOT written, migration runs again next
+    start - and again the start after that, because `_verify_moves()` grades
+    `kept` as well as `moved`. See `_verify_moves()` for the install this was
+    measured on, and for what grading `moved` alone cost on start 2.
+
+    THE THREE RULES ARE `migrate_unit_keys()`'s, SHARED IN CODE rather than
+    restated: `_move_flat_keys()` is rule 1 (never overwrite a namespaced value)
+    and rule 2 (the flat keys are left in the file), `_verify_moves()` is the
+    read-back postcondition, and the marker is written LAST so a crash before it
+    leaves the migration to run again - which rule 1 makes a no-op.
+
+    ⚠️ AND RULE 2 IS KEPT LIVE HERE TOO, WHICH IS THE HALF DEV-02 HAD TO ADD
+    LATER. "Left in place" alone makes the flat history a fossil dated the day
+    the second inverter was declared: an operator who rolls back to a
+    single-FSM add-on would find a `last_completed` months old and recalibrate a
+    pack that does not need it, or - worse - a stale `phase` and a
+    `phase_started` old enough to abort on the 12 h timeout the moment it comes
+    up. So `maintenance.UnitHistory` MIRRORS the first pack's writes back onto
+    the flat keys for as long as the roster has more than one entry. The flat
+    pair is the ROLLBACK VIEW of the history, maintained, exactly as
+    `sign_tripped` is the rollback view of the latch.
+    """
+    rep = MigrationReport()
+    data = getattr(store, "data", {})
+    names = [str(n) for n in roster]
+    rep.from_version = int(data.get(MAINT_SCHEMA_KEY, 0) or 0)
+    if len(names) <= 1:
+        rep.skipped = ("one inverter: the flat maintenance keys ARE that pack's "
+                       "history and are read in place, so there is nothing to "
+                       "move and the store layout does not change")
+        return rep
+    if rep.from_version >= MAINT_SCHEMA_VERSION:
+        rep.skipped = (f"maintenance history already at schema "
+                       f"{rep.from_version}; written by a multi-inverter "
+                       f"version and not touched")
+        return rep
+
+    rep.ran = True
+    _move_flat_keys(store, unit, MAINT_UNIT_KEYS, rep)
+    _verify_moves(store, rep, roster=names)
+
+    # ⚠️ `rep.ok` AND NOTHING ELSE - NOT `ok or kept`, NOT `not lost`, NOT
+    # `not misdirected`. BOTH guards gate this line and each one is pinned by a
+    # check that the other cannot satisfy: section 9c refuses a store whose
+    # failure is `misdirected` with `lost` empty, and section 9d refuses one
+    # whose failure is `lost` with `misdirected` empty AND `kept` non-empty -
+    # the mutation `if rep.ok or rep.kept:` stamps the marker on a migration
+    # that lost a month of history, and until section 9d existed all 19 suites
+    # were green on it.
     if rep.ok:
-        store.set(SCHEMA_KEY, SCHEMA_VERSION)
+        store.set(MAINT_SCHEMA_KEY, MAINT_SCHEMA_VERSION)
     else:
-        # ⚠️ NOT MARKED DONE. A migration that lost state must run again on the
-        # next start rather than be recorded as finished - the one thing worse
-        # than a failed migration is a failed migration a later version skips.
-        _LOG.error("per-inverter store migration LOST STATE and has not been "
-                   "marked done: %s", rep.explain())
+        # ⚠️ NOT MARKED DONE, FOR EITHER FAILURE. The same sentence
+        # `migrate_unit_keys()` carries: the one thing worse than a failed
+        # migration is a failed migration a later version skips. Rule 1 makes
+        # the re-run a no-op where the keys did land, so refusing the marker
+        # costs one repeated pass and buys the difference between a migration
+        # that can still be fixed and one the store says already happened.
+        _LOG.error("per-pack maintenance-history migration did NOT land and "
+                   "has not been marked done: %s", rep.explain())
     return rep
 
 

goodwe_controller/app/validate.py

@@ -74,7 +74,8 @@ NOT READY banner already names the problem, (b) does not apply.
      was overruled. That firmware file's own comment says a lower bound on this
      "belongs in CFG-01, not here" - so here it is.
 
-  5. `sign_disagree_w` at or above `max_w`.  [(a)]
+  5. `sign_disagree_w` - or SAFETY-30's `unit_sign_disagree_w` - at or above
+     `max_w`.  [(a)]
      The arbiter's `supervised` limit bounds every command to +-max_w, and
      SAFETY-02's predicate needs `abs(command) > sign_disagree_w`. Set the
      threshold at or above the ceiling and the command side of the comparison
@@ -131,7 +132,7 @@ NOT READY banner already names the problem, (b) does not apply.
      installed base is on a fixed or static tariff. Refusing to start without a
      price feed would lock those sites out of the add-on entirely.
 
-  8. `peak_soc_floor` at or below `bms_soc_floor`.  [(a)]
+  8. `peak_soc_floor` at or below `bms_soc_floor`, ON ANY UNIT.  [(a)]
      SAFETY-24, and the same shape as 1, 4, 5 and 6. The peak guard can COMMAND
      a discharge to hold a billed quarter down (SAFETY-07), and `peak_soc_floor`
      is the EMS-side bound on how far down it may take the pack. The BMS has its
@@ -141,6 +142,12 @@ NOT READY banner already names the problem, (b) does not apply.
      of binding while its status page reports one - rule 5's "a disarmed guard
      that advertises itself as armed", on the guard whose failure mode is a pack
      that is quietly never available when it is wanted.
+     ⚠️ SAFETY-32: `bms_soc_floor` DESCRIBES A BATTERY, so at a roster it is a
+     per-entry key and this rule grades EVERY unit and NAMES the one it refuses.
+     One number for two packs of different models refuses the wrong config and
+     accepts the wrong one - cut-outs of 10 % and 30 % under `peak_soc_floor: 20`
+     used to start. A missing per-entry cut-out is refused here rather than
+     defaulted: a cut-out invented for a pack IS this defect.
 
      ⚠️ `<=`, NOT `<`, and the equality case is the interesting one. At exactly
      equal, the EMS floor and the pack's cut-out act at the same percent, so the
@@ -234,6 +241,7 @@ from .p1 import is_enabled, validate_meter_options
 from .peakguard import PEAK_OPTION_DEFAULTS
 from .price import PRICE_OPTION_DEFAULTS, PROVIDERS
 from .static_tariff import ProfileInvalid, parse_static_profile
+from .watchdog import SIGN_OPTION_DEFAULTS
 
 _LOG = logging.getLogger("goodwe.config")
 
@@ -384,6 +392,49 @@ def unit_caps_indexed(opts: dict) -> list[tuple[int, float]]:
     return caps
 
 
+def unit_bms_floors_indexed(opts: dict) -> list[tuple[int, object, object]]:
+    """SAFETY-32. Each pack's OWN `bms_soc_floor`, carrying its ROSTER INDEX.
+
+    `(roster index, the raw value as configured, the parsed float or None)`, one
+    triple per unit. The raw value travels too because the refusal quotes what
+    the operator actually wrote, and `!r` of a re-read `opts` key is the wrong
+    string the moment the value came off a roster entry.
+
+    ⚠️ THE INDEX IS THE ROSTER'S, NOT THIS LIST'S, AND THAT IS CFG-04'S BUG
+    WRITTEN DOWN. Rules 5 and 6 fed a CAPS-list index into `_unit_label()`,
+    which indexes the ROSTER, so on a roster whose first entry was malformed the
+    refusal named the WRONG INVERTER - an operator sent to the wrong device by
+    the very message that exists to stop that. `unit_caps_indexed()` exists
+    because of it. This function goes one step further and SKIPS NOTHING: every
+    entry produces a triple, including the ones with no usable floor, because a
+    missing cut-out is itself what rule 8 has to refuse BY NAME. There is
+    therefore no second list for an index to desynchronise from.
+
+    ⚠️ AND NOTHING IS INHERITED. At a declared roster the FLAT `bms_soc_floor` is
+    not read at all - CFG-04's rule, and this card's subject: a pack that borrows
+    a sibling's cut-out is graded against a battery it is not. At the flat shape
+    the flat key IS this one pack's cut-out, defaulted from CFG-03's single
+    mirror exactly as `inverter._flat_bms_floor()` defaults it.
+    """
+    entries = roster_entries(opts)
+    if not entries:
+        return [(0, opts.get("bms_soc_floor"),
+                 _number(opts, "bms_soc_floor",
+                         PEAK_OPTION_DEFAULTS["bms_soc_floor"]))]
+    out: list[tuple[int, object, object]] = []
+    for index, entry in enumerate(entries):
+        if not isinstance(entry, dict):
+            # Rule 9(a)/(b) already refuses a non-dict entry by index, and it
+            # runs FIRST. Naming it twice would hand the operator two sentences
+            # about one line of YAML.
+            continue
+        raw = entry.get("bms_soc_floor")
+        out.append((index, raw,
+                    None if raw is None or raw == ""
+                    else _number(entry, "bms_soc_floor", None)))
+    return out
+
+
 def _unit_label(opts: dict, index: int) -> str:
     """"" at a single-inverter config, " (inverter-b)" at a roster.
 
@@ -741,6 +792,65 @@ def validate_startup_config(opts: dict, dev_min=None, dev_max=None) -> list[str]
             f"sign_disagree_w below {weakest:.0f} W (300 W is the measured "
             "default), or raise max_w.")
 
+    # ⚠️ AND THE SAME RULE FOR SAFETY-30's PER-UNIT THRESHOLD, WHICH IS A
+    # SEPARATE COMPARISON AND THEREFORE A SEPARATE WAY TO DISARM. It grades a
+    # unit's own SHARE against that unit's own measured power, and a share is
+    # bounded by the same `supervised` claim at +-max_w - so a threshold at or
+    # above the ceiling makes the command side unsatisfiable by construction,
+    # exactly as above. Written as its own branch rather than folded into the
+    # one above because the two produce different sentences and an operator
+    # needs to be told WHICH watchdog they have disarmed.
+    #
+    # ⚠️ REFUSED EVEN AT A ROSTER OF ONE, WHERE THIS CHECK IS NOT ARMED. A
+    # config that cannot work is refused when it is WRITTEN, not on the day a
+    # second inverter is added - which is the day nobody is watching the startup
+    # log. This is `meter_max_age_s`'s rule read the same way: the failure it
+    # prevents is silent.
+    #
+    # ⚠️ AND IT READS THE WEAKEST UNIT'S CAP, NOT THE FLAT `max_w` - SAFETY-32
+    # ROUND 5. This branch used to compare against `max_w = _number(opts,
+    # "max_w", 2000.0)`, the FLAT option, which on a roster config is either
+    # ABSENT (so it compared against a 2000 W DEFAULT describing no unit on the
+    # site) or a leftover from a single-inverter install (so it compared against
+    # a number no unit has). Both directions were live and both were measured:
+    #
+    #   a roster of two 2000 W units beside a leftover `max_w: 5000` and
+    #   `unit_sign_disagree_w: 3000` produced ZERO problems - the site STARTED,
+    #   with a per-inverter watchdog reporting `armed` on the status page and
+    #   incapable of tripping on EITHER unit. That is the identical sentence the
+    #   branch above exists to refuse, reached through the branch beside it;
+    #
+    #   and a roster of two 5000 W units with no flat key was REFUSED at
+    #   `unit_sign_disagree_w: 3000` against a 2000 W default the operator never
+    #   wrote, for a threshold both units can reach.
+    #
+    # `unit_caps()`'s own docstring already said why: there is no such thing as
+    # a site rating on the wire, and this watchdog grades ONE UNIT's standing
+    # command against THAT unit's `supervised` bound. So it takes the same
+    # `weakest` the branch above takes, from the same list, and names the same
+    # unit. At a roster of one `unit_caps()` is `[max_w]` and this is the
+    # identical comparison it always was - which is what keeps the shipped
+    # single-inverter refusal byte-for-byte what it was.
+    unit_sign_w = _number(opts, "unit_sign_disagree_w",
+                          float(SIGN_OPTION_DEFAULTS["unit_sign_disagree_w"]))
+    unit_where = (_unit_label(opts, min(caps_ix, key=lambda p: p[1])[0])
+                  if caps_ix else "")
+    if unit_sign_w is None:
+        problems.append(f"unit_sign_disagree_w "
+                        f"{opts.get('unit_sign_disagree_w')!r} is not a number")
+    elif weakest is not None and weakest > 0 and unit_sign_w >= weakest:
+        problems.append(
+            f"unit_sign_disagree_w is {unit_sign_w:.0f} W but max_w{unit_where}"
+            f" is {weakest:.0f} W, so the SAFETY-30 per-inverter sign-disagreement "
+            "watchdog can never fire: no inverter is ever commanded past max_w, "
+            f"and the watchdog needs a command over {unit_sign_w:.0f} W. At two "
+            "inverters that is the guard which catches a unit doing the "
+            "OPPOSITE of what it was told while its sibling covers for it at "
+            "the meter - the case the site-level watchdog is structurally "
+            f"blind to. Lower unit_sign_disagree_w below {weakest:.0f} W "
+            f"({SIGN_OPTION_DEFAULTS['unit_sign_disagree_w']:.0f} W is the "
+            "measured default), or raise max_w.")
+
     # --- 6. a maintenance charge the ceiling can actually accommodate --------
     # ⚠️ Strictly `>`, matching the arbiter exactly. The shaping is
     # `Claim.limit(hi=-charge_w)` against `supervised`'s `lo=-max_w`, so
@@ -768,24 +878,47 @@ def validate_startup_config(opts: dict, dev_min=None, dev_max=None) -> list[str]
     # rule 8 below already makes with `PEAK_OPTION_DEFAULTS` - and
     # test_maintenance.py pins it by reading the shipped file.
     #
-    # ⚠️ AND IT IS GRADED AGAINST THE FIRST UNIT'S CAP, NOT THE SMALLEST AND NOT
-    # THE FLAT ONE. Rule 5 takes the weakest unit because there is a watchdog on
-    # every one of them; there is exactly ONE `Maintenance` in this add-on and
-    # `Controller.maint_unit` names the pack it recalibrates - the first in the
-    # roster, because the cycle's history lives under un-namespaced Store keys.
-    # The shaping claim is filed on THAT actuator and on no other, so the cap
-    # that can contradict it is THAT unit's. Taking `min()` here would refuse a
-    # perfectly workable site - a 5000 W pack being recalibrated beside a 2000 W
-    # sibling the cycle never touches - which is the over-refusal this file's
-    # docstring says gets a validator switched off. When `maint_unit` stops
-    # being `inverters[0]`, this line moves with it.
+    # ⚠️ MAINT-05: IT IS GRADED AGAINST THE WEAKEST UNIT'S CAP NOW, AND THE
+    # COMMENT THIS REPLACES SAID SO WOULD HAPPEN. It read: "there is exactly ONE
+    # `Maintenance` in this add-on and `Controller.maint_unit` names the pack it
+    # recalibrates - the first in the roster, because the cycle's history lives
+    # under un-namespaced Store keys. The shaping claim is filed on THAT
+    # actuator and on no other, so the cap that can contradict it is THAT
+    # unit's. Taking `min()` here would refuse a perfectly workable site - a
+    # 5000 W pack being recalibrated beside a 2000 W sibling the cycle never
+    # touches ... When `maint_unit` stops being `inverters[0]`, this line moves
+    # with it."
+    #
+    # `maint_unit` has stopped being `inverters[0]`. There is one `Maintenance`
+    # per roster entry, each with its own namespaced history, and EVERY pack is
+    # recalibrated on its own schedule - so the sibling "the cycle never
+    # touches" does not exist any more and the over-refusal argument has no
+    # subject. The shaping claim is filed on whichever pack is charging, so the
+    # cap that can contradict it is any cap on the roster, and the one that
+    # contradicts it FIRST is the smallest. This is rule 5's conjunction for
+    # rule 5's reason: a threshold that is fine against the biggest unit says
+    # nothing about the smallest, and the smallest is the one that goes to 0 W.
+    #
+    # ⚠️ THIS REFUSES CONFIGS THAT USED TO START, AND THAT IS THE POINT RATHER
+    # THAN A SIDE EFFECT. `max_w` 5000 on pack A beside 2000 on pack B with
+    # `maintenance_charge_w` 3000 was accepted before this card and was correct
+    # while only pack A was ever recalibrated. Pack B now recalibrates too, and
+    # its cheap-window charge asks the arbiter for `hi = -3000` against a
+    # supervised `lo = -2000`: `lo > hi`, arbiter clause 3, 0 W, contradiction -
+    # a maintenance charge that silently never happens on that pack, which is
+    # exactly what rule 6 exists to refuse at the moment it is WRITTEN. At a
+    # roster of one `min()` over one cap is that cap and the comparison, the
+    # message and the boundary are identical.
     charge_w = _number(opts, "maintenance_charge_w",
                        float(MAINT_OPTION_DEFAULTS["maintenance_charge_w"]))
-    # ⚠️ THE SAME PAIRING AS RULE 5, FOR THE SAME REASON. `caps[0]` is the first
-    # unit WITH A USABLE CAP, which is `inverters[0]` only while every earlier
-    # entry has one; `_unit_label(opts, 0)` was hard-wired to `inverters[0]`
-    # regardless. Both come off the one pair now.
-    maint_index, maint_cap = caps_ix[0] if caps_ix else (0, None)
+    # ⚠️ THE SAME PAIRING AS RULE 5, FOR THE SAME REASON: `unit_caps_indexed()`
+    # SKIPS entries with no usable `max_w`, so a caps-list position is not a
+    # roster position and `_unit_label()` indexes the roster. The index travels
+    # with the number, so the unit the message names is the unit the number came
+    # off - which on a roster whose first entry has a junk `max_w` is not the
+    # same thing at all.
+    maint_index, maint_cap = (min(caps_ix, key=lambda p: p[1]) if caps_ix
+                              else (0, None))
     maint_where = _unit_label(opts, maint_index)
     if charge_w is None:
         problems.append(
@@ -814,28 +947,39 @@ def validate_startup_config(opts: dict, dev_min=None, dev_max=None) -> list[str]
     # in-code mirror, pinned to config.yaml by test_peakguard.py, and main.py
     # reads the same one - exactly as maintenance.MAINT_OPTION_DEFAULTS now is
     # for the `maintenance_*` family.
+    #
+    # ⚠️ SAFETY-32 - GRADED ONCE PER UNIT, AND THE REFUSAL NAMES THE UNIT. One
+    # `bms_soc_floor` for two packs compares the EMS floor against whichever
+    # battery happened to be written into the flat key: a site with cut-outs of
+    # 10 % and 30 % and `peak_soc_floor: 20` STARTED, and the guard would then
+    # command a forced discharge that the second pack's BMS - not the operator's
+    # floor - is the thing that stops. That is the disarmed-backstop state this
+    # rule exists to refuse, reached through the rule itself. So the loop is over
+    # units, and `_unit_label()` is fed the ROSTER index the value came off (see
+    # `unit_bms_floors_indexed()`), because CFG-04 has just fixed the version of
+    # this that named the wrong inverter.
     soc_floor = _number(opts, "peak_soc_floor",
                         PEAK_OPTION_DEFAULTS["peak_soc_floor"])
-    bms_floor = _number(opts, "bms_soc_floor",
-                        PEAK_OPTION_DEFAULTS["bms_soc_floor"])
-    if soc_floor is None or bms_floor is None:
-        problems.append(
-            "peak_soc_floor and bms_soc_floor must both be numbers (got "
-            f"{opts.get('peak_soc_floor')!r} / {opts.get('bms_soc_floor')!r})")
-    elif soc_floor <= bms_floor:
-        problems.append(
-            f"peak_soc_floor is {soc_floor:.0f} % but the pack's own cut-out "
-            f"(bms_soc_floor) is {bms_floor:.0f} %, so SAFETY-07's forced "
-            "discharge has no EMS-side bound at all - THE BMS WINS, and it is "
-            "the pack's protection rather than yours. The guard would report a "
-            "SoC floor on its status page while being incapable of ever being "
-            "the thing that stops the discharge, and the symptom of that is a "
-            "battery repeatedly sitting at the BMS limit with no record of why "
-            f"- which looks exactly like the SoC drift the maintenance cycle "
-            f"exists to correct. Raise peak_soc_floor above {bms_floor:.0f} % "
-            f"(the shipped default is "
-            f"{PEAK_OPTION_DEFAULTS['peak_soc_floor']:.0f} %), or correct "
-            "bms_soc_floor if it does not match this pack.")
+    for _ix, _raw, bms_floor in unit_bms_floors_indexed(opts):
+        where = _unit_label(opts, _ix)
+        if soc_floor is None or bms_floor is None:
+            problems.append(
+                f"peak_soc_floor and bms_soc_floor{where} must both be numbers "
+                f"(got {opts.get('peak_soc_floor')!r} / {_raw!r})")
+        elif soc_floor <= bms_floor:
+            problems.append(
+                f"peak_soc_floor is {soc_floor:.0f} % but the pack's own cut-out "
+                f"(bms_soc_floor{where}) is {bms_floor:.0f} %, so SAFETY-07's forced "
+                "discharge has no EMS-side bound at all - THE BMS WINS, and it is "
+                "the pack's protection rather than yours. The guard would report a "
+                "SoC floor on its status page while being incapable of ever being "
+                "the thing that stops the discharge, and the symptom of that is a "
+                "battery repeatedly sitting at the BMS limit with no record of why "
+                f"- which looks exactly like the SoC drift the maintenance cycle "
+                f"exists to correct. Raise peak_soc_floor above {bms_floor:.0f} % "
+                f"(the shipped default is "
+                f"{PEAK_OPTION_DEFAULTS['peak_soc_floor']:.0f} %), or correct "
+                f"bms_soc_floor{where} if it does not match this pack.")
 
     # --- warnings: real, but not worth refusing over -------------------------
     integ = opts.get("integrator_max_w")

goodwe_controller/app/watchdog.py

@@ -164,11 +164,106 @@ saturation freeze's question (|command - the battery's actual power| > 500 W,
 NOTES.md §4.7 row 4), and the two checks are complementary rather than
 overlapping. Neither substitutes for the other.
 
+⚠️ THAT PARAGRAPH IS ABOUT THE COMMAND-VERSUS-METER COMPARISON, AND SINCE
+SAFETY-30 THAT IS NOT THE ONLY ONE THIS MODULE DRIVES. At a roster of two or
+more each unit is ALSO graded against its own measured power, which is exactly
+the comparison the paragraph says nothing here makes - see the SAFETY-30 section
+below, which says what that does and does not take away from the freeze. At the
+shipped roster of one the paragraph stands unchanged, word for word: the site
+comparison is the only detector armed and the freeze is still the whole answer.
+
 Nor does it see a sign inversion introduced DOWNSTREAM of this add-on - in the
 firmware, on the wire, or in a future driver. Our view stays self-consistent and
 this stays silent. That is a wire-versus-intent problem and it belongs to the
 witness, which is the only vantage point that sees both.
 
+------------------------------------------------------------------------------
+SAFETY-30 - AT TWO INVERTERS THIS IS TWO QUESTIONS, NOT ONE
+
+Everything above was written at a roster of ONE, and at N=1 it is right for a
+reason that stops being true at N=2. `commanded_w` is what THIS controller put
+on THIS actuator. `grid_w` is what the WHOLE SITE is exchanging with the
+network. At one inverter those are two views of one thing - §6.2's identity -
+which is what makes comparing them sound. At two they are not:
+
+  * A FALSE TRIP THE COMPARISON CANNOT AVOID. "Inverter A commanded -1500 W
+    while the meter reads +1500 W" is evidence of nothing at N=2: inverter B may
+    be discharging 3000 W and doing exactly what it was told. No fault anywhere,
+    and the check NOTES.md calls the highest-value entry in the safety table -
+    93 % agreement while healthy against 4 % during runaway, four episodes, zero
+    false positives - fires.
+  * THE CONVERSE, WHICH IS WORSE. A unit genuinely driving the wrong way is
+    MASKED, because the site figure NETS. Commanded +1500 W, measured -1500 W,
+    beside a sibling measured at +1500 W: the meter sits exactly where a healthy
+    site leaves it. This is not a threshold that is set too high - the
+    information is destroyed by the summation, and no tuning recovers it.
+
+So the one comparison becomes two, each asked of the evidence that can answer
+it, and they are DIFFERENT FAULTS rather than one fault at two scales:
+
+  PER UNIT   this unit's standing command against ITS OWN measured power -
+             `Inverter.batt`, which on the live system is holding register
+             0x0518, this inverter's own AC power, per-instance since DEV-02.
+             Conclusion: "THE ACTUATOR I AM DRIVING IS NOT DOING WHAT I TOLD
+             IT". The evidence is about one unit, so the remedy is that unit:
+             the `failsafe` SET, which arbiter.CLAIM_SCOPE has called a UNIT
+             claim from the day it was written, for this exact reason.
+  SITE       the SUM of the standing commands against the meter. Conclusion:
+             "THE SITE AS A WHOLE IS DRIVING THE WRONG WAY" - an inverted gain,
+             an integrator wound past the rail, a strategy pushing the wrong
+             side of the meter. That is the fault the 2026-08-20 capture
+             contains and it is a property of the CONNECTION, so the remedy is
+             the connection: a SITE SET that stops every inverter.
+
+⚠️ AND AT N=1 THERE IS EXACTLY ONE ARMED DETECTOR, WHICH IS THE ONE THAT
+SHIPPED. The two questions above are the SAME SENTENCE at a roster of one: the
+sum of one command is that command, and "the site" and "the unit" name one
+actuator. Arming both would be arming a duplicate - two latches, two banners and
+two chances to spend a false-positive record that was earned by ONE comparison
+on two captured bus days. `Controller.cycle()` therefore hands the single unit's
+watchdog THE METER, under the shipped `sign_disagree_w` / `sign_disagree_s`,
+exactly as before, and never arms the site one. `arm()` below is where that
+choice is written down. Nothing about a one-inverter install changes - not the
+comparison, not the thresholds, not the store keys, not the claim, not the
+banner.
+
+⚠️ WHAT THE PER-UNIT COMPARISON IS NOT. Two guards that overlap without saying
+so is how a false-alarm generator gets built, so both neighbours are named:
+
+  * IT IS NOT SAFETY-31. `circulation.py` compares the units AGAINST EACH OTHER
+    - `sum(|B_i|) - |sum(B_i)|` over MEASURED power - and catches two inverters
+    pushing THROUGH each other, which nets to zero at the meter and about which
+    no single command is evidence at all. This compares each unit against ITS
+    OWN COMMAND. Take two units faithfully executing an allocation that ALLOC-01
+    would refuse: circulating power is the whole loop, and every per-unit sign
+    check here reads perfect agreement. Take one unit doing the exact opposite
+    of its share while its sibling covers for it: this reads the whole
+    disagreement, and circulating power can be zero, because the two measured
+    powers may well agree in sign. Neither is the other with a different
+    threshold.
+  * IT IS NOT THE SATURATION FREEZE, AND THE OVERLAP IS REAL AND BOUNDED. The
+    freeze asks `|command - measured| > saturation_w` (500 W) and concludes "the
+    inverter cannot follow"; it then stops the integrator winding, which is a
+    control action taken every day, on a unit that is merely SATURATED as
+    readily as on one that is reversed, and it neither latches nor parks
+    anything. This asks for OPPOSITION - a sign, on both sides of a threshold,
+    held continuously - which a saturated-but-obedient unit cannot produce at
+    all, and it LATCHES and commands that unit to zero. The freeze is a
+    magnitude test that shapes the loop; this is a direction test that stops an
+    actuator.
+
+⚠️ GATE 1 STILL APPLIES TO BOTH, AND THAT IS A DELIBERATE CHOICE WITH A COST.
+The per-unit comparison could have been armed whenever ANY non-zero command is
+standing - a unit should do what it is told whoever told it - and that would
+catch a wedged inverter during a maintenance charge. It is not, because the
+whole no-false-positive argument above is built on `following`, because the
+freeze already grades that window, and because this card's own warning is that
+SAFETY-02's record was earned at N=1 and must not be spent on a check that fires
+when nothing is wrong. What it costs is written here rather than discovered:
+between the start of a maintenance cycle and its end, a unit reversing against
+its own command is the freeze's to notice, not this module's. Widening it is a
+card, with a rig campaign behind it, not an edit.
+
 ------------------------------------------------------------------------------
 STATE, TRIP AND LATCH
 
@@ -198,6 +293,178 @@ from dataclasses import dataclass
 from datetime import datetime
 
 
+# SAFETY-30. CFG-03: THE ONE IN-CODE COPY OF THE FOUR MULTI-INVERTER DEFAULTS.
+# `config.yaml`'s `options:` block is the source of truth an operator edits and
+# the Supervisor reads, and `schema:` holds the bounds. This dict exists for the
+# reason `circulation.CIRC_OPTION_DEFAULTS` and `maintenance.MAINT_OPTION_DEFAULTS`
+# do - the Dockerfile does not put config.yaml in the image, so the running
+# add-on cannot read it - and `test_watchdog.py` pins the two together by READING
+# config.yaml, so a drift goes red in CI rather than shipping.
+#
+# ⚠️ EVERY KEY HERE IS DEAD AT A ROSTER OF ONE, and that is the point rather
+# than a caveat. `Controller._arm_sign_watchdogs()` arms the single unit's
+# detector against the meter with the SHIPPED `sign_disagree_w` /
+# `sign_disagree_s`, which are read in main.py exactly where they always were
+# and are not in this dict. Nothing below is consulted until a second inverter
+# exists.
+SIGN_OPTION_DEFAULTS = {
+    # --- THE MAGNITUDE GATE, AND IT IS NOT SAFETY-02's 300 W BY ACCIDENT ---
+    # ⚠️ DERIVED FROM `control.Tuning.saturation_w`, WHICH IS 500 W, AND THE
+    # DERIVATION IS THE POINT. This check gates on BOTH sides being over
+    # `min_w`, so the smallest divergence it can ever see between a command and
+    # the power that command produced is `2 * min_w` - the two are OPPOSED, so
+    # the gap is the sum of the magnitudes. At 250 W that is exactly 500 W: the
+    # gap at which `compute()`'s own anti-windup test already concludes THE
+    # INVERTER IS NOT FOLLOWING and freezes the integrator. So this watchdog
+    # cannot fire on a divergence the shipped control law itself calls ordinary
+    # tracking, and the two guards meet at a number rather than overlapping in a
+    # band nobody has drawn.
+    #
+    # MEASURED, on 1890 closed-loop two-inverter runs in which NOTHING is faulty
+    # (every unit executes the share the allocator gave it), swept over the
+    # WHOLE range config.yaml accepts for `gain` float(0.05,1.0) and `slew_w`
+    # int(50,5000), across both families of stimulus that can produce legitimate
+    # opposition. Worst CONTINUOUS opposition between a unit's standing command
+    # and its own 0x0518 power, as a function of this threshold:
+    #
+    #     50 W -> 25.0 s    250 W -> 10.0 s     700 W -> 5.0 s
+    #    150 W -> 15.0 s    300 W -> 10.0 s    1000 W -> 0.0 s
+    #    200 W -> 15.0 s    500 W ->  5.0 s
+    #
+    # ⚠️ AND THE CURVE IS FLAT, WHICH IS THE FINDING THAT DECIDED THE SHAPE OF
+    # THIS PAIR. Unlike the site comparison - where the meter is driven by the
+    # LOAD and a flapping load holds it opposed for minutes - a unit's own power
+    # CHASES ITS OWN COMMAND through a bounded ramp and a first-order lag, so the
+    # opposition can only last as long as the crossing does. Dropping the
+    # threshold from 300 W to 50 W buys 15 s of extra transient and nothing else.
+    # The work is therefore done by the WINDOW below, and this number is set
+    # where the control law already draws its line rather than being tuned.
+    #
+    # The largest opposed magnitude seen anywhere in those 1890 runs is 875 W;
+    # a threshold above that would gate NOTHING legitimate in at all, and is
+    # rejected because it would also be blind to a unit reversing a 900 W share.
+    "unit_sign_disagree_w": 250,
+
+    # --- THE WINDOW -------------------------------------------------------
+    # Three bounds, and the value clears all three.
+    #
+    #   LOWER, from the rig: 10.0 s - two control cycles, the worst CONTINUOUS
+    #   legitimate opposition above 250 W across all 1890 runs. It is a
+    #   reversing site whose second write lands a cycle after the first, both
+    #   units then slewing through zero behind a 10 s lag; nothing is wrong.
+    #
+    #   LOWER, from the firmware: `inverter.LINK_STALE_S`, 30 s. A unit not
+    #   executing its share for less than the firmware's own veto window is a
+    #   LINK fault, and the add-on already answers it - liveness drops the unit
+    #   and the survivors carry its share, with the record saying so. Firing
+    #   inside that window would raise a second, louder alarm about a state
+    #   already being handled, which is the SAFETY-15/SAFETY-19 failure exactly.
+    #
+    #   UPPER, from SAFETY-02 itself: 180 s, the SITE hold. A unit that is not
+    #   doing what it was told must not be parked LATER than the site would be
+    #   parked for the same watts, and unlike the site comparison this one needs
+    #   nothing to net out before it can conclude - the evidence is complete on
+    #   the first cycle it appears.
+    #
+    # 85 s: seventeen control cycles - 8.5x the worst legitimate episode this
+    # rig can produce, 2.8x the firmware's own veto window, and under HALF the
+    # site hold. ⚠️ IT IS DELIBERATELY NOT 60, 75, 90, 120 OR 180. A round
+    # number here would be a guess wearing a measurement's clothes, and the
+    # three bounds above are what the number has to answer to.
+    #
+    # ⚠️ AND INVERTER B DOES NOT PHYSICALLY EXIST. Its ramp and lag in the rig
+    # are inverter A's, COPIED, carrying `measured=False` - SPIKE-04 measures
+    # the real ones. A slower second unit lengthens every crossing above, which
+    # is what the 8.5x buys headroom for. SPIKE-04 is the revisit trigger for
+    # both numbers here, not a reason to move them now.
+    "unit_sign_disagree_s": 85,
+
+    # --- THE SITE COMPARISON AT N>=2, AND IT IS NOT THE SHIPPED PAIR -------
+    # ⚠️ THE THRESHOLD IS CARRIED AND THE WINDOW IS NOT, WHICH IS A FINDING
+    # RATHER THAN A COMPROMISE. Only one of the two arguments behind 300 W /
+    # 180 s depends on how many inverters the site has:
+    #
+    #   THE THRESHOLD is about NOISE. "Near zero the sign is noise, and near
+    #   zero is where a healthy loop LIVES" - the deadband holds the command
+    #   still while the meter wanders either side of `target_grid_w`. The
+    #   deadband is 15 W and P1 is one meter whatever the roster, so nothing
+    #   about that argument moves. 300 W, carried - and the two-inverter
+    #   campaign says the carry is safe rather than merely convenient: in the
+    #   load-reversal family the site figure ever goes opposed AND over 300 W in
+    #   994 of 1050 runs at two inverters and 955 of 1050 at one. The gate opens
+    #   about as often either way; what changed is how LONG it stays open, which
+    #   is the window's problem and not the threshold's.
+    #
+    #   THE WINDOW is about a TRANSIENT, and the transient scales with the site.
+    #   The measured worst legitimate case is a load reversal that the loop
+    #   crosses zero slowly: the command has to travel the whole rail at
+    #   `slew_w` per cycle, and at N>=2 the rail is `Controller.site_max_w` - the
+    #   ROSTER SUM - while `slew_w` is unchanged. So the crossing takes about
+    #   twice as long at two inverters as at one, and that is exactly what the
+    #   rig measures.
+    #
+    # 1890 closed-loop runs in which nothing is faulty, the SAME campaign the
+    # per-unit pair above is read off, run at BOTH rosters as its own control.
+    # Worst CONTINUOUS opposition between the standing command(s) and the meter,
+    # LOAD-REVERSAL family, at a 300 W threshold:
+    #
+    #     N=1 (the shipped comparison)   110.0 s     - inside 180 s, 1.6x
+    #     N=2 (the sum, same stimuli)    215.0 s     - OVER 180 s
+    #
+    # ⚠️ SO CARRYING 180 s WOULD HAVE SHIPPED A FALSE TRIP, and it is a config a
+    # schema-legal install can reach: `gain` 0.05 and `slew_w` 50 are both the
+    # floor of what config.yaml accepts, and a +-2200 W load reversal is an oven
+    # and a heat pump. That is the number this card exists to not assume.
+    #
+    # 325 s: 1.51x the measured worst, well inside the 900 s quarter that bills
+    # the capacity tariff, and still inside every episode the fault day
+    # contains - 2026-08-20's four clusters run 16, 30, 77 and 77 minutes and
+    # its opposed runs reach 1750 s, so a 325 s window fires in all of them.
+    # What it costs is stated rather than hidden: about two and a half extra
+    # minutes of wrong-direction SITE power before the stop, bought to remove a
+    # false trip that a legal configuration produces. ⚠️ DELIBERATELY NOT 300 OR
+    # 360; the measurement is what it answers to.
+    #
+    # ⚠️ AND IT IS NOT THE PER-UNIT WINDOW EITHER, IN THE OTHER DIRECTION. The
+    # per-unit comparison is 85 s because a unit's own power CHASES ITS OWN
+    # COMMAND and the transient is two cycles; the site comparison is 325 s
+    # because the meter is driven by the LOAD and the transient is the loop
+    # crossing zero. Same predicate, two instruments, two windows, both measured.
+    "site_sign_disagree_w": 300,
+    "site_sign_disagree_s": 325,
+}
+
+# ⚠️ A LIMIT OF THE SITE COMPARISON THAT SAFETY-30 DID NOT INTRODUCE AND DOES
+# NOT FIX, RECORDED HERE BECAUSE THE CAMPAIGN FOUND IT AND SILENCE WOULD BE
+# WORSE. The rig's other stimulus family is a FLAPPING LOAD, and a square wave
+# whose half period is about one control cycle opposes the command on EVERY
+# cycle: the command chases the previous half-cycle's meter while the meter has
+# already reversed. Across the 840 flapping runs the worst continuous opposition
+# at a 300 W threshold is 980 s at two inverters - and 1000 s at ONE, which is
+# the whole point. The N=1 arm is the control, and it says this is a property of
+# comparing a lagged command against a meter, not a property of the second
+# inverter. No window inside the 900 s billing quarter survives it, and no
+# threshold up to 1000 W does either.
+#
+# It is not acted on here for three reasons, in order of weight: it is identical
+# at the roster this add-on actually ships, so acting on it would be changing a
+# live safety path on evidence that has nothing to do with this card; a
+# ~6 kW square wave at 0.1 Hz sustained for a quarter of an hour is not a house
+# (the captured healthy day's longest opposed run is 30 s); and the fix is not a
+# number - it wants the detector to notice that the disagreement is ALTERNATING,
+# which is a different check. It is a card, and this comment is its evidence.
+
+# SAFETY-30. THE TWO NOUNS THE TRIP SENTENCE CAN NAME, WRITTEN ONCE. They are
+# constants rather than literals at the call sites because `main.py` arms the
+# detectors and `test_watchdog.py` grades what they say, and a string typed in
+# two places is a string that drifts in one of them - the same argument
+# `store.unit_key()` makes about a key name. `AGAINST_METER` is the wording that
+# has always shipped and must not change: it is what a one-inverter install's
+# `reason` reads, in the store, in the action log and on the status page.
+AGAINST_METER = "a meter reading"
+AGAINST_OWN_POWER = "its own measured power"
+
+
 @dataclass(frozen=True)
 class SignWatchdogConfig:
     """Defaults, not constants - config.yaml carries both (SAFETY-02 AC 5)."""
@@ -206,7 +473,7 @@ class SignWatchdogConfig:
     hold_s: float = 180.0
 
 
-def disagrees(commanded_w, grid_w, following: bool,
+def disagrees(commanded_w, against_w, following: bool,
               cfg: SignWatchdogConfig) -> bool:
     """The predicate, and the ONLY place it is written down.
 
@@ -215,12 +482,23 @@ def disagrees(commanded_w, grid_w, following: bool,
     inline. Two copies of a safety predicate is one copy plus a future
     disagreement about what the check means, and the copy that drifts is always
     the one in the harness that grades it.
+
+    ⚠️ `against_w` WAS CALLED `grid_w`, AND THE RENAME IS THE WHOLE OF SAFETY-30
+    IN ONE PARAMETER. The arithmetic is unchanged - opposed signs, neither side
+    near zero, and the loop actually driving - but WHAT the command is opposed
+    TO is now a property of the caller rather than a fact about this function.
+    The site comparison hands it the meter; each unit's own comparison hands it
+    that unit's measured power (register 0x0518). One predicate, two callers,
+    for `loop_is_following()`'s reason: a safety predicate written down twice is
+    one copy plus a future disagreement about what the check means. The
+    parameter is positional at every call site in both repositories, so the
+    rename moves no caller.
     """
     return (following
-            and commanded_w is not None and grid_w is not None
-            and commanded_w * grid_w < 0
+            and commanded_w is not None and against_w is not None
+            and commanded_w * against_w < 0
             and abs(commanded_w) > cfg.min_w
-            and abs(grid_w) > cfg.min_w)
+            and abs(against_w) > cfg.min_w)
 
 
 def loop_is_following(resolution, loop_asked) -> bool:
@@ -272,7 +550,17 @@ def loop_is_following(resolution, loop_asked) -> bool:
 
 
 class SignWatchdog:
-    """One instance per actuator, evaluated once per control cycle.
+    """One instance per COMPARISON, evaluated once per control cycle.
+
+    ⚠️ SINCE SAFETY-30 THAT IS NOT THE SAME SENTENCE AS "one per actuator",
+    which is what this docstring used to say. There is one instance per
+    inverter - `Inverter.sign_wd`, latched under that unit's own store keys -
+    AND one for the site, `Controller.site_sign_wd`, latched under flat ones.
+    The two ask different questions of different evidence and reach different
+    remedies; `arm()` below is where an instance is told which of them it is.
+    At a roster of one they would be the same question, so only the per-unit
+    instance is ever armed there - see the SAFETY-30 section of the module
+    docstring.
 
     ⚠️ NOT SHARED ACROSS CADENCES. `disagree_since` is a wall-clock start time,
     so driving one instance from two loops at two rates would have each one
@@ -282,8 +570,14 @@ class SignWatchdog:
     """
 
     def __init__(self, cfg: SignWatchdogConfig | None = None, *,
-                 tripped: bool = False, reason: str = ""):
+                 tripped: bool = False, reason: str = "",
+                 against: str = AGAINST_METER):
         self.cfg = cfg or SignWatchdogConfig()
+        # SAFETY-30. WHAT THE COMMAND IS BEING COMPARED AGAINST, as the operator
+        # reads it in the trip sentence. The default is the meter, which is what
+        # every trip this add-on has ever taken was compared against and is what
+        # keeps a one-inverter install's `reason` string byte-identical.
+        self.against = str(against)
         self.disagree_since: datetime | None = None
         # ⚠️ RESTORED FROM THE STORE, NOT DEFAULTED. A latch that lives only in
         # memory is not a latch: a container restart clears it, and the add-on
@@ -292,8 +586,31 @@ class SignWatchdog:
         self.tripped = bool(tripped)
         self.reason = reason if self.tripped else ""
 
+    # -- which of SAFETY-30's two questions this instance is asking ----------
+    def arm(self, cfg: SignWatchdogConfig, against: str) -> None:
+        """Set the thresholds and the subject of the comparison. Nothing else.
+
+        ⚠️ IT IS CONFIGURATION, NOT STATE, AND IT DELIBERATELY DOES NOT TOUCH
+        `disagree_since`, `tripped` OR `reason`. A detector that cleared its own
+        window when it was told what it was comparing against would lose the
+        episode it is in the middle of, every cycle, and would never accumulate
+        a hold at all - the check would be permanently green and nothing would
+        say so.
+
+        ⚠️ AND IT IS RE-STATED EVERY CYCLE RATHER THAN FROZEN AT CONSTRUCTION.
+        The answer is a property of the ROSTER and of nothing else, and
+        `Controller.cycle()` already re-reads the roster every cycle for
+        liveness and for the allocation. Freezing it in `__init__` would also
+        make it wrong for any caller that appends an inverter afterwards, which
+        is exactly how a two-inverter fixture is built in this repository - and
+        a fixture that silently got the one-inverter comparison would grade
+        nothing while looking green.
+        """
+        self.cfg = cfg
+        self.against = str(against)
+
     # -- the whole check, and it is a pure function of its arguments ----------
-    def evaluate(self, commanded_w, grid_w, following: bool,
+    def evaluate(self, commanded_w, against_w, following: bool,
                  now: datetime) -> bool:
         """Advance the detector one cycle. Returns True while tripped.
 
@@ -301,15 +618,23 @@ class SignWatchdog:
                        after deadband, slew, quantisation and the arbiter. Not
                        the optimizer's raw wish: the wire carries the emitted
                        value, so anything else can disagree with the meter for a
-                       reason that is not a fault.
-        `grid_w`       net P1, + = importing. None when the meter is stale or
-                       missing, which CLEARS the timer rather than accumulating
-                       - a frozen figure held against a moving command would
-                       otherwise manufacture a disagreement in about three
-                       minutes, on top of SAFETY-01 having already commanded
-                       zero for the same cause.
+                       reason that is not a fault. For a per-unit instance it is
+                       THAT UNIT'S share of it; for the site instance it is the
+                       roster sum.
+        `against_w`    what the command is graded against, and SAFETY-30 made
+                       this a decision rather than a fact: net P1 for the SITE
+                       comparison (+ = importing), and this unit's own measured
+                       power for the PER-UNIT one (+ = discharging, register
+                       0x0518). Both carry the same sign convention as the
+                       command, which is why one predicate serves both. None
+                       when the reading is stale or missing, which CLEARS the
+                       timer rather than accumulating - a frozen figure held
+                       against a moving command would otherwise manufacture a
+                       disagreement in about three minutes, on top of SAFETY-01
+                       having already commanded zero for the same cause.
         `following`    did the grid-following loop actually win this cycle? See
-                       gate 1 in the module docstring.
+                       gate 1 in the module docstring, and the SAFETY-30 section
+                       on why the per-unit comparison keeps it.
         `now`          the caller's clock, so this stays testable and replayable.
         """
         if self.tripped:
@@ -317,7 +642,7 @@ class SignWatchdog:
             return True
 
         cfg = self.cfg
-        if not disagrees(commanded_w, grid_w, following, cfg):
+        if not disagrees(commanded_w, against_w, following, cfg):
             self.disagree_since = None
             return False
 
@@ -330,9 +655,17 @@ class SignWatchdog:
             return False
 
         self.tripped = True
+        # ⚠️ `self.against` IS THE ONLY THING SAFETY-30 CHANGED IN THIS STRING,
+        # and it defaults to the wording that shipped - "a meter reading" - so a
+        # one-inverter install's `reason` is byte-identical to what an operator
+        # (and test_watchdog.py) has always read. A per-unit instance names what
+        # it actually compared instead, because "commanded +1500 W against a
+        # meter reading -1500 W" would be a false statement about which
+        # instrument produced the evidence, on the one line somebody woken by
+        # this alarm reads first.
         self.reason = (
-            f"SIGN_DISAGREEMENT: commanded {commanded_w:+.0f} W against a meter "
-            f"reading {grid_w:+.0f} W for {held:.0f} s continuously "
+            f"SIGN_DISAGREEMENT: commanded {commanded_w:+.0f} W against "
+            f"{self.against} {against_w:+.0f} W for {held:.0f} s continuously "
             f"(limits: both over {cfg.min_w:.0f} W, held over {cfg.hold_s:.0f} s)"
         )
         return True

goodwe_controller/config.yaml

@@ -56,23 +56,58 @@ options:
   #       soc_entity: sensor.goodwe_a_soc
   #       batt_entity: sensor.goodwe_a_battery_power
   #       max_w: 5000
+  #       bms_soc_floor: 10
   #     - name: inverter-b
   #       setpoint_entity: number.goodwe_b_setpoint_w
   #       soc_entity: sensor.goodwe_b_soc
   #       batt_entity: sensor.goodwe_b_battery_power
-  #       max_w: 3000
+  #       max_w: 4600
+  #       bms_soc_floor: 10
+  #
+  # ⚠️ THE 4600 IS INVERTER B'S DATASHEET FIGURE, NOT A MEASUREMENT AND NOT A
+  # PLACEHOLDER. SPIKE-04a (2026-09-04) identified B as a GoodWe GW5048D-ES,
+  # serial `55048ESU22AW0992`, whose datasheet gives Max. Charging / Discharging
+  # Power as 4600 W; nothing will be connected to its MPPTs, so the battery path
+  # is the ceiling and the 5000 VA AC nameplate is not. The 3000 W this example
+  # used to carry was a guess made before the unit was known. ⚠️ It is STILL not
+  # measured: the datasheet prints 4600 W under BATTERY INPUT DATA, so it may be
+  # DC, and `max_w` is graded against 0x0518 - the inverter's own AC power. If it
+  # is DC the AC ceiling is nearer 4324 W. SAFETY-20's `min(max_w, dev_max)` is
+  # what settles that once the device range is read; see goodwe/NOTES.md §1.1b,
+  # which records both readings and picks neither.
+  #
+  # ⚠️ AND THE `bms_soc_floor: 10` BESIDE IT IS FROM THE SAME PLACE - THE
+  # DATASHEET AND THE SERIAL, NOT A PACK ANYONE HAS READ. No second inverter
+  # exists on this site (SPIKE-04, ≈ 2026-09-12 for B); the cut-out must be read
+  # off the pack that is actually installed (register 0x0560 on the reference
+  # unit) before that pack is commanded. It reads 10 here because pack A reads
+  # 10, which makes this example DEGENERATE ON THE FLOOR AXIS: a site whose two
+  # packs floor at the SAME percent cannot separate `min`, `max`, first or last
+  # over the roster, and every SAFETY-32 fixture that DIFFERS the floors differs
+  # it for exactly that reason. Do not "simplify" those fixtures to match this
+  # example.
   #
   # ⚠️ NOTHING IS INHERITED FROM THE FLAT KEYS BY AN ENTRY THAT OMITS IT, and
   # that is the whole of DEV-02 restated as a config rule. `soc`, `batt` and the
   # rating became per-instance precisely so that no inverter is ever graded
   # against a sibling's numbers; an entry silently inheriting `soc_entity` would
-  # rebuild that defect in YAML. `name`, `setpoint_entity` and `max_w` are
-  # therefore REQUIRED of every entry (the schema below enforces it, and
+  # rebuild that defect in YAML. `name`, `setpoint_entity`, `max_w` and
+  # `bms_soc_floor` are therefore REQUIRED of every entry (the schema below enforces it, and
   # validate.py rule 9 enforces it again for the callers the Supervisor never
   # sees). An omitted `soc_entity`/`batt_entity` means THIS UNIT HAS NO SUCH
   # SENSOR - `inputs_ok` already stops the site for that - and never means
   # "use the other one's".
   #
+  # ⚠️ SAFETY-32: `bms_soc_floor` IS PER ENTRY AND IS REQUIRED, NOT DEFAULTED.
+  # It is the percent THAT pack's own BMS cuts out at, and two batteries of
+  # different models do not share one. A default would be this add-on inventing
+  # a cut-out for a battery it has not been told about, which is the same defect
+  # as inheriting one: the shallower pack gets protected by a floor that belongs
+  # to the deeper pack, and validate.py rule 8 - which refuses a peak_soc_floor
+  # the BMS would win against - would be grading the wrong number. Rule 8 grades
+  # EVERY entry and names the one it refuses. The flat key below is unchanged and
+  # is the single-inverter shape's own cut-out.
+  #
   # ⚠️ AND THE FLAT `setpoint_entity` MUST THEN BE BLANK. Leaving it set while a
   # roster is declared is two config sites aimed at actuators, one of which
   # nothing writes; if it names an entity an entry also names, SAFETY-28's
@@ -167,6 +202,62 @@ options:
   sign_disagree_w: 300
   sign_disagree_s: 180
 
+  # --- SAFETY-30 per-inverter sign-disagreement watchdog ---------------------
+  # ⚠️ A DIFFERENT COMPARISON, NOT A SECOND COPY OF THE ONE ABOVE, AND IT IS
+  # ARMED ONLY WHEN THIS SITE HAS MORE THAN ONE INVERTER. The pair above grades
+  # the SITE's total command against P1 - one meter, one grid connection. This
+  # pair grades EACH inverter's own share against THAT inverter's own measured
+  # power (holding register 0x0518), and it catches the fault the meter cannot
+  # answer for once there are two units: a unit doing the opposite of what it
+  # was told, while its sibling covers for it and P1 reads exactly what a
+  # healthy site reads.
+  #
+  # At ONE inverter these two keys do nothing at all. §6.2's identity makes the
+  # site's total command that inverter's command, so the pair above is already
+  # asking this question of the same actuator, and arming a second detector
+  # would be two latches and two banners for one fault.
+  #
+  # Both defaults are measured on admin/goodwe's two-master rig, over 1890
+  # closed-loop runs in which nothing is faulty, swept across the whole range
+  # this file accepts for `gain` and `slew_w`. 250 W is `saturation_w / 2`, so
+  # the smallest divergence this can see is exactly the gap at which the control
+  # law itself already says the inverter is not following; 85 s is 8.5x the
+  # worst legitimate opposition those runs produce (10.0 s), 2.8x the firmware's
+  # own 30 s link veto, and under half the 180 s the site check takes. See
+  # app/watchdog.py's SIGN_OPTION_DEFAULTS and
+  # DECISIONS/SAFETY-30-per-unit-sign.md.
+  #
+  # ⚠️ unit_sign_disagree_w must stay BELOW max_w, for sign_disagree_w's reason:
+  # a share is bounded to +-max_w, so a threshold at or above it can never fire.
+  # validate.py refuses that config.
+  unit_sign_disagree_w: 250
+  unit_sign_disagree_s: 85
+
+  # --- SAFETY-30 SITE sign-disagreement watchdog -----------------------------
+  # ⚠️ THE SAME COMPARISON `sign_disagree_*` MAKES, AT THE ROSTER WHERE IT STOPS
+  # BEING A STATEMENT ABOUT ONE INVERTER. At two or more units the SUM of the
+  # standing commands is graded against P1, and the trip stops the whole site -
+  # what it catches is the site being asked for the wrong thing, which every
+  # per-inverter check is blind to because each unit is faithfully executing its
+  # share. At ONE inverter these two keys do nothing: `sign_disagree_*` above is
+  # already asking this of the same actuator.
+  #
+  # The threshold is the shipped 300 W, CARRIED, and the reason is that the
+  # argument behind it does not depend on the roster: near zero the sign is
+  # noise, the deadband is 15 W, and there is one P1 whatever is behind it.
+  #
+  # ⚠️ THE WINDOW IS NOT CARRIED, AND THAT IS A MEASUREMENT. The transient 180 s
+  # exists to reject is the loop crossing zero after a load reversal, and the
+  # command has to travel the whole SITE rail at `slew_w` per cycle - a rail
+  # that is the roster sum while `slew_w` is not. On 1890 two-inverter runs in
+  # which nothing is faulty, the worst continuous legitimate opposition at 300 W
+  # is 110 s at one inverter and 215 s at two: carrying 180 s across would have
+  # shipped a false trip reachable at `gain: 0.05`, `slew_w: 50` and an ordinary
+  # 2.2 kW load reversal. 325 s is 1.51x that, inside the 900 s billing quarter,
+  # and still inside every episode of the 2026-08-20 fault day.
+  site_sign_disagree_w: 300
+  site_sign_disagree_s: 325
+
   # --- SAFETY-18 frozen-but-fresh meter detector -----------------------------
   # Seconds the meter may keep answering with an UNCHANGED measurement while the
   # command stands at the rail before the add-on raises an alarm. Both halves
@@ -299,6 +390,13 @@ options:
   # DoD 90. It is here so validate.py can REFUSE a peak_soc_floor at or below
   # it: such a floor can never bind, the BMS wins, and the status page would
   # report an energy backstop that does not exist.
+  #
+  # ⚠️ SAFETY-32 - THIS KEY DESCRIBES ONE PACK, so it is the SINGLE-INVERTER
+  # shape's number and nothing more. Declare an `inverters:` roster and each
+  # entry states its own `bms_soc_floor`; this one is not read at all and is
+  # not inherited by any entry. Rule 8 then grades every unit separately, which
+  # is the whole of that card: one cut-out for two packs of different models
+  # protects the shallower pack with the deeper pack's number.
   bms_soc_floor: 10
 
   # ⚠️ THE OLD PRICE PATH, AND IT IS NOT DEAD YET. These two entities are the
@@ -444,6 +542,7 @@ schema:
     - name: str
       setpoint_entity: str
       max_w: int(100,5000)
+      bms_soc_floor: int(0,50)
       soc_entity: str?
       batt_entity: str?
       batt_invert: bool?
@@ -487,6 +586,33 @@ schema:
   sign_disagree_w: int(50,5000)
   sign_disagree_s: int(30,1800)
 
+  # SAFETY-30. No "off" value, for SAFETY-02's reason.
+  #
+  # ⚠️ THE THRESHOLD'S RANGE IS THE SAME AS THE SITE CHECK'S ON PURPOSE, AND THE
+  # WINDOW'S FLOOR IS NOT. On the two-inverter rig the per-unit comparison's
+  # worst continuous legitimate opposition is nearly FLAT in the threshold -
+  # 25.0 s at 50 W, 15.0 s at 150 W, 10.0 s at 250 W, 5.0 s at 500 W - because a
+  # unit's own power chases its own command rather than the load. So no value in
+  # this range puts the guard outside the regime it can be answered for, and the
+  # floor does not have to be a knee the way `circulation_w`'s does.
+  unit_sign_disagree_w: int(50,5000)
+  # The floor of 30 is `inverter.LINK_STALE_S`, the firmware's own veto window:
+  # below it, a unit not executing its share is a LINK fault the add-on already
+  # answers by dropping the unit from the split, and firing inside that window
+  # would be a second, louder alarm about a state that is already being handled.
+  unit_sign_disagree_s: int(30,1800)
+
+  # SAFETY-30's SITE pair. Same ranges as `sign_disagree_*`, because it is the
+  # same comparison asked of the site total - and no "off" value, for the same
+  # reason. ⚠️ THERE IS DELIBERATELY NO validate.py RULE PAIRING
+  # site_sign_disagree_w WITH max_w, unlike the other two thresholds: the bound
+  # on the SITE command is `Controller.site_max_w`, the ROSTER SUM, and
+  # `validate_startup_config()` is handed options rather than a roster. A rule
+  # written against max_w would refuse a perfectly reachable two-inverter
+  # configuration the day CFG-04 lands. The schema range is the bound here.
+  site_sign_disagree_w: int(50,5000)
+  site_sign_disagree_s: int(30,1800)
+
   # No "off" value, for SAFETY-02's reason: disarming a safety alarm is not a
   # supported configuration. ⚠️ THE FLOOR OF 60 IS NOT A RECOMMENDATION. At 300 s
   # the slowest a HEALTHY meter's 1 Wh register can tick and still look unchanged

goodwe_controller/test_arbiter.py

@@ -165,21 +165,29 @@ SITE_SOURCES = {
                    "identically zero for any one unit - there is no fact of "
                    "the matter about which of two opposing inverters is the "
                    "offender, so there is nothing to file it per-unit on",
+    # SAFETY-30, and it is the entry most easily confused with `failsafe`
+    # below, because the same three words describe both.
+    "sign_site": "the SUM of every standing command against the meter. The "
+                 "offending number is the SITE target, which no single "
+                 "inverter owns - parking one of them would leave the rest "
+                 "pushing the same way with the guard satisfied",
 }
 UNIT_SOURCES = {
     "device":     "one inverter's own reported min/max rating",
     "supervised": "max_w caps what may be WRITTEN to one setpoint entity",
-    "failsafe":   "site-wide evidence, per-unit remedy: SAFETY-02 grades the "
-                  "command THIS controller put on THIS actuator",
+    "failsafe":   "per-unit evidence, per-unit remedy: since SAFETY-30 it "
+                  "grades THIS unit's command against THIS unit's OWN "
+                  "measured power (0x0518), so what it learns is about one "
+                  "device. The site-wide half moved to `sign_site`",
     "maintenance": "the monthly recalibration is a pack procedure",
 }
 for _src, _why in sorted(SITE_SOURCES.items()):
     print(f"    site  {_src:<11} {_why}")
 for _src, _why in sorted(UNIT_SOURCES.items()):
     print(f"    unit  {_src:<11} {_why}")
-check("all NINE claim sources are classified, and only those nine",
+check("all TEN claim sources are classified, and only those ten",
       set(CLAIM_SCOPE) == set(SITE_SOURCES) | set(UNIT_SOURCES))
-check("the five that are about the GRID CONNECTION are site claims",
+check("the six that are about the GRID CONNECTION are site claims",
       {s for s, v in CLAIM_SCOPE.items() if v == SCOPE_SITE} == set(SITE_SOURCES))
 check("the four that are about a PACK, a DEVICE or one write are per-inverter",
       {s for s, v in CLAIM_SCOPE.items() if v == SCOPE_UNIT} == set(UNIT_SOURCES))
@@ -200,7 +208,7 @@ check("with one inverter every source routes to the site, per-inverter ones "
 check("...and with none configured at all, likewise",
       all(actuator_for(s, "inverter-a", ()) == ACTUATOR_SITE for s in CLAIM_SCOPE))
 check("with TWO the identity breaks and the split becomes real: the four "
-      "per-inverter sources route to the unit this controller drives, the five "
+      "per-inverter sources route to the unit this controller drives, the six "
       "site sources stay on the grid connection",
       {s: actuator_for(s, "inverter-a", TWO) for s in CLAIM_SCOPE}
       == {**{s: "inverter-a" for s in UNIT_SOURCES},

goodwe_controller/test_inverter.py

@@ -192,10 +192,27 @@ class Hass:
 
 A_ENT = {"setpoint_entity": "number.sp_a", "soc_entity": "sensor.soc_a",
          "batt_entity": "sensor.batt_a"}
-# ⚠️ A HYPOTHESIS, NOT A MEASUREMENT. Inverter B does not exist; 3000 W is
-# chosen only because it DIFFERS from A's 5000 W, which is what lets a check
-# tell B's number from A's. SPIKE-04 measures the real one.
+# ⚠️ SPIKE-04a (2026-09-04) ESTABLISHED B'S REAL RATING FROM ITS DATASHEET:
+# a GoodWe GW5048D-ES, serial `55048ESU22AW0992`, 4600 W on the battery path
+# (goodwe/NOTES.md §1.1b). THIS FIXTURE DELIBERATELY DOES NOT USE IT. 3000 W is
+# kept because the only property these checks consume is that B's number is
+# UNMISTAKABLY not A's, and the real pair is 5000/4600 - a 400 W gap, narrower
+# than `slew_w` and inside the range where an off-by-one-unit bug and ordinary
+# control arithmetic produce indistinguishable numbers. Collapsing the fixture
+# onto the real values would cost these checks their separating power to buy
+# realism they do not use. So: ILLUSTRATIVE, NOT B'S RATING, and not a
+# placeholder for it either - the placeholder era ended on 2026-09-04.
 B_RATING_HYPOTHESIS = 3000.0
+# SAFETY-32. Each pack's OWN BMS cut-out percent, now a REQUIRED roster key.
+# ⚠️ 10 IS MEASURED (register 0x0560 on the reference GW5000S-BP and the shipped
+# `bms_soc_floor` default); 15 IS SIMULATED - there is no second pack. They
+# DIFFER on purpose: a fixture whose two per-unit numbers are equal cannot show
+# a check that read the wrong unit's, which is MUTATION.md habit 1 and the defect
+# that has now bitten four consecutive cards. Both sit below the shipped
+# `peak_soc_floor` of 20, so rule 8 accepts this roster.
+A_BMS_FLOOR = 10
+B_BMS_FLOOR_SIM = 15
+
 A_RATING = 5000.0
 
 BASE_OPTS = dict(A_ENT, meter_source="off", meter_entity="sensor.grid",
@@ -572,6 +589,58 @@ check("...and a store that only fails to persist a DEFAULT is not reported as "
       "dismiss before the real one arrives",
       migrate_unit_keys(LosesTheLatch(), UNIT_A).ok is True)
 
+# ⚠️ AND THE MARKER GATE, ON A STORE WHERE SOMETHING WAS ALREADY NAMESPACED.
+# FOUND BY MAINT-05's ROUND-3 BATTERY, AND IT WAS A SURVIVOR OF ALL 19 FILES.
+# `migrate_unit_keys()` gates `SCHEMA_KEY` on `rep.ok`, and every check above
+# reaches that gate on a store where NOTHING was already under `unit/...` - so
+# `rep.kept` is empty in all of them, and the mutation
+#
+#     if rep.ok:            ->    if rep.ok or rep.kept:
+#
+# stamped the marker on a migration that DROPPED THE LATCH, with the whole suite
+# green. That is a disarmed safety guard plus a store that says the migration
+# finished, so the next start skips the one thing that would have re-armed it -
+# DEV-02's own worst case with the paper trail filed under "done".
+#
+# The fixture is the one shape that separates them: `sign_reason` ALREADY
+# namespaced (so `kept` is non-empty and the mutant's `or` fires) and the
+# `sign_tripped: true` write dropped on the floor (so `lost` is non-empty and
+# `ok` is False). `migrate_unit_keys()` is passed no roster, so `misdirected`
+# cannot be what refuses this - it is the `lost` guard or nothing.
+_kept_reason = "SIGN_DISAGREEMENT: already namespaced by a later version"
+_broken_kept = LosesTheLatch({
+    "sign_tripped": True,
+    "sign_reason": "SIGN_DISAGREEMENT: measured",
+    unit_key(UNIT_A, "sign_reason"): _kept_reason})
+_rep_kept = migrate_unit_keys(_broken_kept, UNIT_A)
+print(f"    a dropped latch write beside an already-namespaced reason: "
+      f"{_rep_kept.explain()}")
+print(f"    lost={len(_rep_kept.lost)} kept={len(_rep_kept.kept)} "
+      f"ok={_rep_kept.ok}; marker "
+      f"{_broken_kept.data.get(SCHEMA_KEY, '<<absent>>')!r}")
+check("PRECONDITION: this fixture reaches the marker gate through the `lost` "
+      "guard with a NON-EMPTY `kept`, which is the combination none of the "
+      "checks above has. One key dropped, one key left alone, `ok` False. If a "
+      "later edit empties `kept` here the mutation goes invisible again, so the "
+      "shape is asserted rather than assumed",
+      _rep_kept.ran and len(_rep_kept.lost) == 1 and len(_rep_kept.kept) == 1
+      and not _rep_kept.ok
+      and _broken_kept.data[unit_key(UNIT_A, "sign_reason")] == _kept_reason)
+check("⚠️ AND THE STORE IS NOT MARKED MIGRATED. The key that was ALREADY there "
+      "does not buy a marker for the latch that was not: stamping here would "
+      "tell the next start that the migration which dropped `sign_tripped` had "
+      "finished, so it would never run again and the guard would stay off on an "
+      "install whose file still reads `sign_tripped: true`",
+      SCHEMA_KEY not in _broken_kept.data)
+check("...and the counterfactual, one property different: the SAME store shape "
+      "on a store that does not drop writes IS marked migrated. So the refusal "
+      "above is about the latch that was lost, not about `kept` being non-empty",
+      migrate_unit_keys(
+          MemStore({"sign_tripped": True,
+                    "sign_reason": "SIGN_DISAGREEMENT: measured",
+                    unit_key(UNIT_A, "sign_reason"): _kept_reason}),
+          UNIT_A).ok is True)
+
 _broken2 = LosesTheLatch({"sign_tripped": True,
                           "sign_reason": "SIGN_DISAGREEMENT: measured"})
 _c2d, _ = built(store=_broken2)
@@ -609,28 +678,40 @@ print("3. AC 3: a trip on B does not latch A, and it survives a restart")
 # inverters get a share of ONE site target, so an ordinary wrong-way loop trips
 # BOTH - correctly, because both really are pushing the wrong way. The state
 # this part is about is the one the flat store keys made unreachable: ONE
-# inverter standing on a command that opposes the meter while the other does
-# not. It is reached the way the field would reach it - inverter B's writes are
-# being REJECTED, so `inv.last_write` stays where it was while A's tracks the
-# loop - and `standing` is what SAFETY-02 grades, per unit.
+# inverter standing on a command that opposes what it is graded against while
+# the other does not. It is reached the way the field would reach it - inverter
+# B's writes are being REJECTED, so `inv.last_write` stays where it was while
+# A's tracks the loop - and `standing` is what SAFETY-02 grades, per unit.
+#
+# ⚠️ SAFETY-30 CHANGED WHAT "OPPOSES" MEANS HERE, AND THE FIXTURE HAD TO MOVE
+# WITH IT - WHICH IS THIS CHECK DOING ITS JOB. It used to stand B on a -1500 W
+# charge against a +2000 W METER and let the site figure convict it. At a roster
+# of two that comparison is exactly the defect SAFETY-30 removes: the meter
+# answers for BOTH inverters, so it is evidence about neither. B is now
+# convicted by its OWN power sensor - standing on a -1500 W charge while
+# `sensor.batt_b` measures it DISCHARGING at +1500 W, which is one inverter
+# doing the opposite of what it was told and is a fact about that inverter
+# alone. Inverter A's own reading stays at 0 W, so A is never gated in.
 #
 # ⚠️ AND THE REJECTION WINDOW IS WHY THIS RUNS IN THREE SECONDS AND NOT THIRTY.
-# `sign_disagree_s` is 1 s here; `LINK_STALE_S` is 30 s, so B is still LIVE for
-# the whole run and the trip is graded before liveness has anything to say.
+# `unit_sign_disagree_s` is 1 s here; `LINK_STALE_S` is 30 s, so B is still LIVE
+# for the whole run and the trip is graded before liveness has anything to say.
 def _tripping_site():
     hass = Hass(
         {"sensor.grid": 2000.0, "sensor.soc_a": 50.0, "sensor.batt_a": 0.0,
-         "sensor.soc_b": 50.0, "sensor.batt_b": 0.0},
+         "sensor.soc_b": 50.0, "sensor.batt_b": 1500.0},
         {"number.sp_a": (-A_RATING, A_RATING, 1.0),
          "number.sp_b": (-B_RATING_HYPOTHESIS, B_RATING_HYPOTHESIS, 1.0)},
         reject={"number.sp_b"})
     store = MemStore()
     c, _ = built({"sign_disagree_w": 300, "sign_disagree_s": 1,
+                  "unit_sign_disagree_w": 300, "unit_sign_disagree_s": 1,
                   "auto_start": True}, store=store, hass=hass)
     c.inverters.append(inverter_b(store, sign_cfg=c.inverters[0].sign_wd.cfg))
     c.auto = True
-    # B is standing on a charge command while the house imports 2000 W. Its
-    # writes are rejected, so this is what it goes on standing at.
+    # B is standing on a charge command while its own sensor says it is
+    # discharging. Its writes are rejected, so this is what it goes on
+    # standing at.
     c.inverters[1].last_write = -1500.0
     return c, hass, store
 
@@ -681,10 +762,12 @@ check("⚠️ ...AND THE TRIP IS MIRRORED ONTO THE PRE-DEV-02 FLAT PAIR, WHICH I
 
 # THE RESTART, which is the whole reason the latch is persisted at all.
 _h3b = Hass({"sensor.grid": 2000.0, "sensor.soc_a": 50.0,
-             "sensor.batt_a": 0.0, "sensor.soc_b": 50.0, "sensor.batt_b": 0.0},
+             "sensor.batt_a": 0.0, "sensor.soc_b": 50.0,
+             "sensor.batt_b": 1500.0},
             {"number.sp_a": (-A_RATING, A_RATING, 1.0),
              "number.sp_b": (-B_RATING_HYPOTHESIS, B_RATING_HYPOTHESIS, 1.0)})
 _c3r, _ = built({"sign_disagree_w": 300, "sign_disagree_s": 1,
+                 "unit_sign_disagree_w": 300, "unit_sign_disagree_s": 1,
                  "auto_start": True}, store=_s3, hass=_h3b)
 _c3r.inverters.append(inverter_b(_s3,
                                  sign_cfg=_c3r.inverters[0].sign_wd.cfg))
@@ -703,12 +786,32 @@ check("a brand new Controller against the same store comes back with B "
 check("...and A goes on carrying the site target across the restart too",
       _c3r.allocation.shares.get("inverter-a") == _c3r.allocation.sum_w
       and _c3r.allocation.sum_w != 0)
-check("...and the status page shows BOTH rows - one latched, one armed. A "
-      "single row would have to pick one inverter to describe, and 'the sign "
-      "watchdog is armed' printed over a latched sibling is the page lying",
-      len([r for r in _c3r.checks() if "sign watchdog" in r["text"]]) == 1
+check("...and the status page shows BOTH per-inverter rows - one latched, one "
+      "armed. A single row would have to pick one inverter to describe, and "
+      "'the sign watchdog is armed' printed over a latched sibling is the page "
+      "lying",
+      len([r for r in _c3r.checks()
+           if r["text"].startswith("sign watchdog [")]) == 1
       and len([r for r in _c3r.checks()
-               if "SIGN-DISAGREEMENT FAILSAFE" in r["text"]]) == 1)
+               if "SIGN-DISAGREEMENT FAILSAFE [" in r["text"]]) == 1)
+check("⚠️ AND SAFETY-30's SITE ROW IS THERE BESIDE THEM, EXACTLY ONCE. At two "
+      "inverters there are three sign questions on this page and they are three "
+      "rows: is A doing what it was told, is B, and is the SITE'S TOTAL "
+      "pointing the right way. The last one is not derivable from the other "
+      "two - both units can execute their shares perfectly while the number "
+      "they were given is itself wrong - so a page that showed only the "
+      "per-unit rows would be advertising two guards where the install has "
+      "three",
+      len([r for r in _c3r.checks()
+           if r["text"].startswith("site sign watchdog:")]) == 1)
+check("...and the per-inverter rows now say WHAT the unit was compared "
+      "against, because at a roster of two it is no longer the meter. A row "
+      "still reading 'the command opposed the meter' would send an operator to "
+      "look at P1 for a fault that is visible only on one inverter's own power "
+      "sensor",
+      all("its own measured power" in r["text"] for r in _c3r.checks()
+          if r["text"].startswith(("sign watchdog [",
+                                   "SIGN-DISAGREEMENT FAILSAFE ["))))
 
 # The operator's clear, which is one button and has to reach every latch.
 run(_c3r.handle_action("auto_toggle"))          # -> stopped
@@ -728,8 +831,9 @@ check("...and the same press cleared the flat rollback view, so a downgrade "
       "- and neither does the next start, which now reads that key on EVERY "
       "boot rather than only on the one that migrates",
       _s3.data["sign_tripped"] is False and _s3.data["sign_reason"] == "")
-_c3r2, _ = built({"sign_disagree_w": 300, "sign_disagree_s": 1}, store=_s3,
-                 hass=_h3b)
+_c3r2, _ = built({"sign_disagree_w": 300, "sign_disagree_s": 1,
+                  "unit_sign_disagree_w": 300, "unit_sign_disagree_s": 1},
+                 store=_s3, hass=_h3b)
 check("⚠️ AND THE RESTART AFTER THAT CLEAR COMES UP UNLATCHED. This is the "
       "'re-arm forever' hazard the every-start guard would have if the clear "
       "did not reach both spellings, driven through the real button and a real "
@@ -1324,13 +1428,16 @@ _TWO_OPTS.pop("batt_entity")
 _TWO_OPTS["inverters"] = [
     dict(name="inverter-a", setpoint_entity="number.sp_a",
          soc_entity="sensor.soc_a", batt_entity="sensor.batt_a",
-         max_w=A_RATING),
+         max_w=A_RATING, bms_soc_floor=A_BMS_FLOOR),
     # ⚠️ 3000 W IS A HYPOTHESIS AND IS MARKED AS ONE WHEREVER IT APPEARS. It is
     # here because it DIFFERS from A's 5000 W, which is the only property any
     # check below uses. SPIKE-04 measures the real device.
+    # ⚠️ AND SO IS ITS 15 % CUT-OUT (SAFETY-32). There is no second pack to read
+    # register 0x0560 off; the value is SIMULATED, and it differs from A's 10 %
+    # for MUTATION.md habit 1's reason and no other.
     dict(name="inverter-b", setpoint_entity="number.sp_b",
          soc_entity="sensor.soc_b", batt_entity="sensor.batt_b",
-         max_w=B_RATING_HYPOTHESIS),
+         max_w=B_RATING_HYPOTHESIS, bms_soc_floor=B_BMS_FLOOR_SIM),
 ]
 _c9, _hass9 = built(_TWO_OPTS)
 print(f"    the roster built from `inverters:` -> {[repr(i) for i in _c9.inverters]}")
@@ -1406,14 +1513,19 @@ _decl9 = dict(ln.strip().lstrip("- ").split(": ", 1)
               for ln in _entry9.splitlines()[1:] if ": " in ln)
 print(f"    the shipped per-entry schema: {_decl9}")
 check("...and the schema half declares it as a LIST OF OBJECTS whose `name`, "
-      "`setpoint_entity` and `max_w` are REQUIRED - no trailing `?` - so the "
-      "Supervisor refuses a half-written entry before the add-on sees it, and "
-      "an entry can never fall back on a sibling's rating because there is no "
-      "entry without one",
-      set(_decl9) == {"name", "setpoint_entity", "max_w", "soc_entity",
-                      "batt_entity", "batt_invert"}
+      "`setpoint_entity`, `max_w` and `bms_soc_floor` are REQUIRED - no "
+      "trailing `?` - so the Supervisor refuses a half-written entry before the "
+      "add-on sees it, and an entry can never fall back on a sibling's rating "
+      "or on a sibling's BMS cut-out because there is no entry without one. "
+      "⚠️ SAFETY-32 ADDED THE FOURTH ONE AS REQUIRED RATHER THAN AS "
+      "`int(0,50)?`, and the difference is the whole of that card: an optional "
+      "per-pack cut-out is a DEFAULTED one, and a default is one number for "
+      "two batteries - the shallower pack protected by the deeper pack's BMS",
+      set(_decl9) == {"name", "setpoint_entity", "max_w", "bms_soc_floor",
+                      "soc_entity", "batt_entity", "batt_invert"}
       and not any(_decl9[k].endswith("?")
-                  for k in ("name", "setpoint_entity", "max_w"))
+                  for k in ("name", "setpoint_entity", "max_w",
+                            "bms_soc_floor"))
       and all(_decl9[k].endswith("?")
               for k in ("soc_entity", "batt_entity", "batt_invert")))
 
@@ -1495,9 +1607,11 @@ check("...and two inverters on two DIFFERENT entities are accepted, because "
       "that is the configuration this card exists for and a validator that "
       "fires on the legitimate case gets switched off",
       validate_startup_config(_opts(setpoint_entity="", inverters=[
-          dict(name="inverter-a", setpoint_entity="number.sp_a", max_w=A_RATING),
+          dict(name="inverter-a", setpoint_entity="number.sp_a", max_w=A_RATING,
+               bms_soc_floor=A_BMS_FLOOR),
           dict(name="inverter-b", setpoint_entity="number.sp_b",
-               max_w=B_RATING_HYPOTHESIS)])) == [])
+               max_w=B_RATING_HYPOTHESIS,
+               bms_soc_floor=B_BMS_FLOOR_SIM)])) == [])
 
 # --- the rest of rule 9 -------------------------------------------------------
 _p_flat = validate_startup_config(_opts(inverters=[
@@ -1594,11 +1708,28 @@ check("⚠️ RULE 5 NAMES THE UNIT THE NUMBER CAME OFF. The weakest cap is "
       "is inverter-B. The refusal says inverter-c",
       _why(_p_skew, "max_w (inverter-c) is 1000 W")
       and not _why(_p_skew, "max_w (inverter-b) is 1000 W"))
-check("⚠️ RULE 6 TOO. `caps[0]` is the first entry WITH a usable cap - "
-      "inverter-b - while `_unit_label(opts, 0)` was hard-wired to inverter-a, "
-      "so the maintenance refusal quoted B's rating under A's name",
-      _why(_p_skew, "max_w (inverter-b) is 2000 W")
-      and not _why(_p_skew, "max_w (inverter-a) is 2000 W"))
+# ⚠️ MAINT-05 MOVED RULE 6's SUBJECT, AND THIS CHECK MOVED WITH IT. It used to
+# assert the refusal named inverter-b, because rule 6 graded the FIRST entry
+# with a usable cap - `Controller.maint_unit` was `inverters[0]` and the shaping
+# claim could be filed nowhere else. There is now one `Maintenance` per roster
+# entry, so the shaping claim can be filed on ANY pack and the cap that
+# contradicts it first is the SMALLEST - inverter-c's 1000 W. The property this
+# check exists for is unchanged and is the reason it is still here: the index
+# has to travel with the cap, or the message quotes one unit's rating under
+# another unit's name.
+check("⚠️ RULE 6 TOO, AND SINCE MAINT-05 IT IS THE WEAKEST PACK IT NAMES. "
+      "inverter-c's 1000 W sits at caps index 1, and `_unit_label(opts, 1)` is "
+      "inverter-B - so a caps index leaking into a roster index would say "
+      "inverter-b here. The refusal says inverter-c, with inverter-c's number",
+      _why(_p_skew, "maintenance_charge_w is 3000 W but max_w (inverter-c) is "
+                    "1000 W")
+      and not _why(_p_skew, "maintenance_charge_w is 3000 W but max_w "
+                            "(inverter-b)"))
+check("...and it is no longer graded against the FIRST usable cap at all, which "
+      "is the behaviour change MAINT-05 makes here: inverter-b's 2000 W was the "
+      "pre-card answer and appears nowhere in the maintenance refusal now",
+      not _why(_p_skew, "maintenance_charge_w is 3000 W but max_w "
+                        "(inverter-b) is 2000 W"))
 check("...and the counterfactual, so this is not a check that would pass on a "
       "validator that named nothing: the OLD expressions still evaluate, on "
       "this same config, to the wrong names",
@@ -2097,6 +2228,16 @@ MUTATIONS = [
     ("validate.py allowing two roster entries with the SAME name",
      "part 10: 'two entries with the SAME name are refused' - one Store "
      "namespace for two units is DEV-02's defect rebuilt in YAML"),
+    ("`migrate_unit_keys()`'s marker gate `if rep.ok:` -> `if rep.ok or "
+     "rep.kept:` - the schema marker stamped on a migration that DROPPED THE "
+     "LATCH, because one key happened to be namespaced already",
+     "part 10: 'AND THE STORE IS NOT MARKED MIGRATED'. ⚠️ THIS ONE SURVIVED "
+     "ALL 19 FILES and was found by MAINT-05's round-3 battery, not by this "
+     "card: every migration check here reached the gate on a store where "
+     "NOTHING was namespaced yet, so `rep.kept` was empty in all of them and "
+     "the `or` never fired. Stamping there disarms the latch AND tells the "
+     "next start the migration finished, so it never re-arms - DEV-02's own "
+     "worst case filed under 'done'"),
 ]
 # ⚠️ FOUR OF THESE SURVIVED THE FIRST TIME AND THE CHECKS WERE CHANGED, NOT THE
 # TABLE - two in DEV-02, two more in CFG-04, and the CFG-04 pair are recorded in
@@ -2115,7 +2256,7 @@ for what, caught in MUTATIONS:
 check(f"all {len(MUTATIONS)} defeats of this card are caught by a named check "
       f"above. A guard whose checks survive its own defeat is not a guard, and "
       f"this repository has three separate records of exactly that",
-      len(MUTATIONS) == 34)
+      len(MUTATIONS) == 35)
 
 
 print()

goodwe_controller/test_validate.py

@@ -204,12 +204,17 @@ check("...and the refusal names the actuator and every writer aimed at it, so "
 _PER_INVERTER_OK = {k: v for k, v in base().items() if k != "setpoint_entity"}
 _PER_INVERTER_OK["inverters"] = [
     {"name": "inverter-a", "setpoint_entity": "number.goodwe_setpoint",
-     "max_w": 2000},
+     "max_w": 2000, "bms_soc_floor": 10},
     # ⚠️ INVERTER B DOES NOT EXIST. 2000 W here is the SHIPPED default repeated,
     # not a rating measured off a second device - SPIKE-04 measures that when
     # there is one to measure. Nothing in this file depends on the number.
+    # ⚠️ SAFETY-32 made `bms_soc_floor` a REQUIRED per-entry key, and the two
+    # values here DIFFER (10 / 15) rather than being copied. 15 % is SIMULATED -
+    # there is no second pack to read a cut-out off - and the reason it is not
+    # simply 10 is MUTATION.md habit 1: a fixture whose two per-unit numbers are
+    # equal cannot show a check reading the wrong unit's.
     {"name": "inverter-b", "setpoint_entity": "number.goodwe_setpoint_b",
-     "max_w": 2000}]
+     "max_w": 2000, "bms_soc_floor": 15}]
 check("two inverters with their own entities are ACCEPTED - one writer each, "
       "no shared actuator, and that is CFG-04's whole purpose",
       validate_startup_config(_PER_INVERTER_OK, -5000, 5000) == [])
@@ -353,6 +358,64 @@ check("...and it is refused with maintenance_enabled off, as shipped",
           base(max_w=2000, maintenance_charge_w=2500,
                maintenance_enabled=False), -5000, 5000), "THE CEILING WINS"))
 
+# ⚠️ MAINT-05: AND AT TWO PACKS IT IS THE WEAKEST CAP, NOT THE FIRST ENTRY'S.
+# The comment this replaces in validate.py said, in as many words, that it took
+# the FIRST unit's cap because there was exactly one `Maintenance` and it could
+# only ever recalibrate `inverters[0]` - and that "when `maint_unit` stops being
+# `inverters[0]`, this line moves with it". It has. Every pack is now
+# recalibrated on its own schedule, so the shaping claim can be filed on any of
+# them and the cap that contradicts it first is the smallest.
+#
+# ⚠️ THE FIXTURE'S TWO CAPS ARE DELIBERATELY DIFFERENT AND THE RATE SITS
+# BETWEEN THEM. Two equal caps could not tell "grades the first" from "grades
+# the smallest", and a rate above both would be refused by either rule - so
+# 3000 W against caps of 5000 W and 2000 W is the one arrangement where the old
+# answer and the new one disagree. ⚠️ 2000 W for the second pack is a
+# HYPOTHESIS: inverter B does not physically exist and nothing here is measured
+# from one.
+def _roster(a_max, b_max, **over):
+    o = base(**over)
+    for k in ("setpoint_entity", "soc_entity", "batt_entity", "max_w"):
+        o.pop(k, None)
+    o["inverters"] = [
+        dict(name="inverter-a", setpoint_entity="number.sp_a",
+             soc_entity="sensor.soc_a", batt_entity="sensor.batt_a",
+             max_w=a_max),
+        dict(name="inverter-b", setpoint_entity="number.sp_b",
+             soc_entity="sensor.soc_b", batt_entity="sensor.batt_b",
+             max_w=b_max)]
+    return o
+
+
+_two = _roster(5000, 2000, maintenance_charge_w=3000, maintenance_discharge_w=1000,
+               sign_disagree_w=300, unit_sign_disagree_w=300)
+_p_two = [p for p in validate_startup_config(_two, -5000, 5000)
+          if "maintenance_charge_w" in p]
+print(f"    5000 W and 2000 W packs, charge 3000 W -> {_p_two}")
+check("⚠️ A CHARGE RATE THE SMALLEST PACK CANNOT TAKE IS REFUSED, even though "
+      "the FIRST pack could. Before MAINT-05 the second pack was never "
+      "recalibrated, so its cap could not contradict anything; it is now, and "
+      "its cheap-window charge would ask for hi=-3000 against lo=-2000 - the "
+      "`lo > hi` contradiction, 0 W, on the pack nobody was watching",
+      len(_p_two) == 1 and why(_p_two, "max_w (inverter-b) is 2000 W"))
+check("...and the message names the pack the number came off, not the first "
+      "entry - an operator told to lower a number on the wrong device is being "
+      "sent to the wrong device by the message that exists to stop that",
+      not why(_p_two, "max_w (inverter-a)"))
+check("...while the same rate against two packs that can both take it starts, "
+      "so this is a comparison and not a blanket refusal of rosters",
+      [p for p in validate_startup_config(
+          _roster(5000, 4000, maintenance_charge_w=3000,
+                  maintenance_discharge_w=1000, sign_disagree_w=300,
+                  unit_sign_disagree_w=300), -5000, 5000)
+       if "maintenance_charge_w" in p] == [])
+check("...and at a ROSTER OF ONE `min()` over one cap is that cap: the same "
+      "message, the same boundary, the same refusal the flat config always got",
+      why(validate_startup_config(base(max_w=2000, maintenance_charge_w=2500),
+                                  -5000, 5000), "THE CEILING WINS")
+      and validate_startup_config(base(max_w=2000, maintenance_charge_w=2000),
+                                  -5000, 5000) == [])
+
 # ⚠️ THE SIBLING, CHECKED RATHER THAN ASSUMED. The shaping is NOT symmetric:
 # drain files Claim.set at P_MAINTENANCE and owns the actuator, so the ceiling
 # CLAMPS it (+2000 W) instead of contradicting it, and the arbiter's explain()
@@ -523,9 +586,26 @@ check("...and the message explains it would read as a trip, not a mistake",
 # check below.
 check("the smallest sane max_w still starts",
       validate_startup_config(base(max_w=1, sign_disagree_w=0.5,
+                                   unit_sign_disagree_w=0.5,
                                    maintenance_charge_w=1), -5000, 5000) == [])
 check("...but that same 1 W ceiling with the DEFAULT sign threshold is refused",
       why(validate_startup_config(base(max_w=1), -5000, 5000), "never fire"))
+# ⚠️ SAFETY-30 ADDED A SECOND WAY TO DISARM A SIGN WATCHDOG BY CONFIGURATION,
+# and the fixture above now has to name BOTH thresholds - which is the check
+# doing its job rather than a bookkeeping edit. `unit_sign_disagree_w` bounds a
+# unit's own SHARE, which the same `supervised` claim holds at +-max_w, so at or
+# above the ceiling it is unsatisfiable by construction exactly as
+# `sign_disagree_w` is. Refused even at the shipped roster of one, where this
+# watchdog is not armed: a config that cannot work is refused when it is
+# written, not on the day somebody adds a second inverter.
+check("...and the per-inverter threshold has the same rule, named separately "
+      "so an operator is told WHICH watchdog they disarmed",
+      why(validate_startup_config(base(max_w=1, sign_disagree_w=0.5,
+                                       maintenance_charge_w=1), -5000, 5000),
+          "unit_sign_disagree_w"))
+check("...and one watt under that ceiling is reachable, so accepted",
+      validate_startup_config(base(max_w=2000, unit_sign_disagree_w=1999),
+                              -5000, 5000) == [])
 
 # --------------------------------------------------------------------------- #
 print("what must NOT be refused - a validator that fires on trivia gets deleted")
@@ -753,6 +833,339 @@ check("...and MaintConfig's dataclass defaults carry no numeric literal either "
       not re.search(r"^\s*(charge_w|discharge_w):\s*float\s*=\s*-?\d",
                     _srcs["maintenance.py"], re.M))
 
+# --------------------------------------------------------------------------- #
+print("SPIKE-04a: inverter B's documented rating, and the provenance beside it")
+# ⚠️ WHY A DOCUMENTATION CARD LEFT A CHECK BEHIND. Inverter B's rating was
+# written down in THREE places - config.yaml's commented roster example,
+# DOCS.md's, and tools/supervisor_shape_probe.py's "a blob an operator would
+# POST" - as 3000 W, a figure nobody had measured, read off a datasheet, or
+# sourced at all. It was wrong: SPIKE-04a (2026-09-04) identified inverter B
+# from the owner's model number and serial as a GoodWe GW5048D-ES rated 4600 W
+# on the battery path. Three transcriptions of one number drifting away from
+# the truth together is this repo's recurring defect in its plainest form, and
+# correcting them once WITHOUT PINNING THEM buys nothing: the next editor
+# touches one of the three and the other two stay wrong, silently.
+#
+# ⚠️ AND THE PROVENANCE IS GRADED, NOT ONLY THE VALUE, WHICH IS THE HALF THAT
+# MATTERS. 4600 W IS FROM A DATASHEET AND IS NOT A MEASUREMENT. The datasheet
+# prints it under BATTERY INPUT DATA, so it is probably DC, while `max_w` is
+# graded against 0x0518 - the inverter's own AC power - which would put the AC
+# ceiling nearer 4324 W. An editor who tidies that caveat away leaves a paper
+# number reading as an observation of hardware that does not exist yet, and
+# THAT is the failure this card exists to prevent. So the caveat is what these
+# checks reach for; a check that graded 4600 alone would watch the value and
+# miss the lie.
+#
+# ⚠️ WHAT THIS DELIBERATELY DOES NOT GRADE: the test fixtures. Several files
+# build a second inverter at 3000 W on purpose, because the property those
+# checks consume is that B's rating is UNMISTAKABLY not A's, and the real pair
+# (5000/4600) is a 400 W gap - narrower than `slew_w`. Those are illustrative,
+# they say so at the line, and they are not statements of B's rating. The three
+# sites below are the ones that ARE such statements.
+_HERE = os.path.dirname(os.path.abspath(__file__))
+_docs_txt = open(os.path.join(_HERE, "DOCS.md"), encoding="utf-8").read()
+_probe_txt = open(os.path.join(_HERE, "tools", "supervisor_shape_probe.py"),
+                  encoding="utf-8").read()
+
+# One regex per site, each anchored on something only THAT site has, so a match
+# cannot silently wander into a neighbouring blob. The probe's anchor is the
+# operator blob's real entity name: its section D refusal entries use
+# `number.b` and one of them carries a deliberate 9999, so an anchor loose
+# enough to reach them would grade the wrong dict and still look green.
+_b_rating = {
+    "config.yaml": re.search(
+        r"#\s+- name: inverter-b\b(?:[^\n]*\n)*?\s*#\s+max_w:\s*(\d+)", _txt),
+    "DOCS.md": re.search(
+        r"^  - name: inverter-b\b(?:[^\n]*\n)*?^    max_w:\s*(\d+)",
+        _docs_txt, re.M),
+    "supervisor_shape_probe.py": re.search(
+        r'"name": "inverter-b",\s*"setpoint_entity": "number\.goodwe_b_setpoint_w"'
+        r'(?:[^}]*?)"max_w":\s*(\d+)', _probe_txt),
+}
+check(f"all three documented rosters were FOUND "
+      f"{ {k: bool(v) for k, v in _b_rating.items()} } - a regex that matches "
+      "nothing is a green check that read nothing",
+      all(_b_rating.values()))
+# ⚠️ THE `len(...) == 3` IS NOT DECORATION. Without it a regex that stopped
+# matching would shrink this dict, and "all the values agree" would go GREEN on
+# the two sites that were still read - a check whose data no longer reaches the
+# failure it names, which is the exact fault this whole section is about.
+_b_vals = {k: int(m.group(1)) for k, m in _b_rating.items() if m}
+check(f"...and all three state the SAME rating for inverter B {_b_vals}",
+      len(_b_vals) == 3 and len(set(_b_vals.values())) == 1)
+check(f"...and it is 4600 W - the GW5048D-ES datasheet figure, not the 3000 W "
+      f"hypothesis that predated the unit being identified {_b_vals}",
+      len(_b_vals) == 3 and set(_b_vals.values()) == {4600})
+
+# The provenance triple. Each site must still say WHERE 4600 came from, that it
+# is graded against an AC register, and what the DC-vs-AC ambiguity does to it -
+# the last carried by the 4324 W figure, which is the one marker that cannot
+# survive a careless "it's 4600 now, done" rewrite.
+#
+# ⚠️ SEARCHED IN A WINDOW AROUND THE ROSTER, NOT OVER THE WHOLE FILE, AND THE
+# FIRST VERSION OF THIS CHECK DID THE LATTER. `0x0518` already appeared once in
+# config.yaml and once in DOCS.md before this card, for unrelated reasons - so a
+# whole-file `"0x0518" in src` was GREEN on those two files no matter what
+# happened next to inverter B's rating. That is a check whose data cannot reach
+# the failure it names, in the same file that exists to complain about them.
+# The window is the text either side of the roster match, so deleting the caveat
+# beside the number goes red even where the same word survives elsewhere.
+_sites = {"config.yaml": _txt, "DOCS.md": _docs_txt,
+          "supervisor_shape_probe.py": _probe_txt}
+_MARKERS = ("datasheet", "0x0518", "4324")
+_prov = {}
+for _name, _src in _sites.items():
+    _m = _b_rating[_name]
+    _win = _src[max(0, _m.start() - 2500):_m.end() + 2500].lower() if _m else ""
+    _prov[_name] = {k: k in _win for k in _MARKERS}
+for _name, _got in sorted(_prov.items()):
+    check(f"{_name} still carries B's provenance WITHIN 2500 chars of the "
+          f"roster - the datasheet it came from, the 0x0518 it is graded "
+          f"against, and the 4324 W the DC reading would make it {_got}",
+          all(_got.values()))
+
+
+# --------------------------------------------------------------------------- #
+print("rules 5, 5b and 6 on a ROSTER SET - the smallest cap at EVERY position")
+# ⚠️ SAFETY-32 ROUND 5, AND THE REASON THIS SECTION EXISTS AT ALL. PR #48 was
+# rejected on FOUR MUTATIONS THAT SURVIVED A 19/19 GREEN SUITE, all of them in
+# the three lines above:
+#
+#   validate.py:779  min(caps)   -> min(caps[:1])                  +4 B  GREEN
+#   validate.py:779  min(caps)   -> min(caps[-1:])                 +5 B  GREEN
+#   validate.py:778  caps = [cap for _, cap in caps_ix[:1]]        +4 B  GREEN
+#   validate.py:893  min(caps_ix[-1:], key=lambda p: p[1])         +5 B  GREEN
+#
+# NONE of them is an equivalent mutant: each one deletes a safety refusal.
+# `max_w: 5000` on inverter-a beside `2000` on inverter-b with
+# `sign_disagree_w: 3000` is refused by the shipped validator with "the
+# SAFETY-02 sign-disagreement watchdog can never fire" and ACCEPTED by every one
+# of the four - the site starts, with a watchdog that reports `armed` on the
+# status page and cannot trip on pack B. That is rule 5's own failure mode
+# reached THROUGH rule 5.
+#
+# WHY THE OLD FIXTURE COULD NOT SEE IT, which is the whole lesson and is now
+# MUTATION.md habit 10: the fixture was ONE roster, `[5000, 2000]`. On that
+# roster `min` and `last` and `[-1:]` and `[:2]` all answer 2000. A fixture of N
+# members cannot separate two reductions that agree on N members, and adding a
+# member moves the hole instead of closing it.
+#
+# SO THE FIXTURE IS A SET, NOT A ROSTER: the smallest cap at EVERY POSITION of
+# rosters of width 2, 3, 4 and 5. `[:1]` needs width 2 to be caught, `[:2]`
+# needs 3, `[:3]` needs 4 - the sweep is what terminates the regress that a
+# single fixture cannot. And the FAMILY TABLE below is COMPUTED AND PRINTED, so
+# a member no fixture kills shows up as a NAMED SURVIVOR rather than as an
+# absence from a list nobody re-derived.
+#
+# ⚠️ THE DIFFERING CAPS ARE A FIXTURE PROPERTY AND MUST STAY ONE. In production
+# both packs are expected to floor at 10 % and SPIKE-04a puts inverter B at
+# 4600 W against A's 5000 W - a 400 W gap, narrower than slew_w. A fixture built
+# on the real pair would be degenerate on exactly the axis these checks measure.
+# ⚠️ AND INVERTER B DOES NOT PHYSICALLY EXIST (SPIKE-04, ~2026-09-12). Nothing
+# in this section is measured off a second unit; the rosters are simulated and
+# are labelled so.
+_BIG, _SMALL = 5000.0, 2000.0
+_RATE = 3000.0                     # strictly between them, which is the point
+
+
+def _roster_n(caps, **over):
+    """A roster of len(caps) units, unit-0..unit-N, each with its own cap."""
+    o = base(**over)
+    for k in ("setpoint_entity", "soc_entity", "batt_entity", "max_w"):
+        o.pop(k, None)
+    o["inverters"] = [
+        dict(name=f"unit-{i}", setpoint_entity=f"number.sp{i}",
+             soc_entity=f"sensor.soc{i}", batt_entity=f"sensor.batt{i}",
+             max_w=c, bms_soc_floor=10)
+        for i, c in enumerate(caps)]
+    o.update(over)
+    return o
+
+
+# The three rules that reduce over `unit_caps_indexed()`, each as
+# (option name, the token its refusal is recognised by).
+_CAP_RULES = (
+    ("sign_disagree_w", "SAFETY-02 sign-disagreement"),
+    ("unit_sign_disagree_w", "SAFETY-30 per-inverter sign-disagreement"),
+    ("maintenance_charge_w", "maintenance charge in a cheap window"),
+)
+_QUIET = {"sign_disagree_w": 300.0, "unit_sign_disagree_w": 300.0,
+          "maintenance_charge_w": 1000.0, "maintenance_discharge_w": 1000.0}
+
+# THE SET: width x position of the one small cap.
+_SET = []
+for _n in (2, 3, 4, 5):
+    for _pos in range(_n):
+        _caps = [_BIG] * _n
+        _caps[_pos] = _SMALL
+        _SET.append((_n, _pos, tuple(_caps)))
+
+# habit 1: the PRECONDITION as its own check. If a later edit makes the caps
+# equal, this goes red rather than the claims below going quietly vacuous.
+check("PRECONDITION: every fixture in the set has a UNIQUE smallest cap, that "
+      "cap is at the stated position, and the rate under test sits STRICTLY "
+      "between it and the others - two equal caps could not tell `min` from "
+      "`first` from `last` and the whole set would be measuring nothing",
+      len(_SET) == 14
+      and all(c[p] == _SMALL and c.count(_SMALL) == 1 and _SMALL < _RATE < _BIG
+              for _n2, p, c in _SET)
+      and {p for _n2, p, _c in _SET} == {0, 1, 2, 3, 4})
+
+
+def _verdict(opts, token):
+    """(refused?, the unit the message NAMES, the cap it QUOTES)."""
+    ps = [p for p in validate_startup_config(opts, -5000, 5000) if token in p]
+    if not ps:
+        return (False, None, None)
+    m = re.search(r"max_w \(([^)]+)\) is (\d+) W", ps[0])
+    return (True, m.group(1) if m else None, float(m.group(2)) if m else None)
+
+
+# --- the shipped answer, on every member of the set -------------------------
+_rows = []
+for _opt, _token in _CAP_RULES:
+    for _n, _pos, _caps in _SET:
+        _o = _roster_n(list(_caps), **dict(_QUIET, **{_opt: _RATE}))
+        _rows.append((_opt, _n, _pos, _caps, _verdict(_o, _token)))
+print(f"    {len(_CAP_RULES)} rules x {len(_SET)} rosters = {len(_rows)} runs "
+      f"of the REAL validator (rate {_RATE:.0f} W, caps {_SMALL:.0f}/"
+      f"{_BIG:.0f} W, simulated rosters - inverter B does not exist):")
+for _opt in sorted({r[0] for r in _rows}):
+    _got = [(r[1], r[2], r[4]) for r in _rows if r[0] == _opt]
+    print(f"      {_opt:<22} " + " ".join(
+        f"n{n}p{p}:{'REFUSE ' + str(u) if ok else 'starts'}"
+        for n, p, (ok, u, _c) in _got))
+
+check("⚠️ EVERY ONE OF THE THREE CAP RULES REFUSES ON EVERY MEMBER OF THE SET, "
+      "and NAMES THE UNIT THE NUMBER CAME OFF - at width 2, 3, 4 and 5, with "
+      "the small cap at every position in turn. A reduction that answered "
+      "`first`, `last`, `[:1]`, `[-1:]`, `[:2]` or `[:-1]` instead of `min` "
+      "misses at least one member of this set, which is what four hand-run "
+      "mutations against a 19/19 green suite proved the single `[5000, 2000]` "
+      "fixture could not do",
+      all(ok and unit == f"unit-{p}" and cap == _SMALL
+          for _o2, _n2, p, _c2, (ok, unit, cap) in _rows))
+
+check("...and it is a COMPARISON, not a blanket refusal of rosters: the same "
+      "three rates against a roster whose caps can ALL take them start "
+      "cleanly, at every width. Without this the line above is satisfied by a "
+      "validator that refuses everything (habit 2 in the negative direction)",
+      all(_verdict(_roster_n([_BIG] * n, **dict(_QUIET, **{opt: _RATE})),
+                   tok) == (False, None, None)
+          for opt, tok in _CAP_RULES for n in (1, 2, 3, 4, 5)))
+
+# --- the POSITIONAL-SLICE FAMILY TABLE, computed and printed ----------------
+# ⚠️ COMPUTED, NOT LISTED. Each candidate is applied to the SAME caps the
+# validator saw, and its (unit, cap) answer is compared against what the shipped
+# validator actually said on that roster. A candidate that agrees everywhere is
+# a reduction this fixture set CANNOT distinguish from the shipped one, and it
+# is printed as a SURVIVOR rather than left out of a table nobody re-derived.
+def _pick(caps, fn):
+    ix = list(enumerate(caps))
+    try:
+        i, c = fn(ix)
+    except (ValueError, IndexError):
+        return None
+    return (f"unit-{i}", c)
+
+
+_FAMILY = {
+    "min (SHIPPED)":  lambda ix: min(ix, key=lambda p: p[1]),
+    "max":            lambda ix: max(ix, key=lambda p: p[1]),
+    "first":          lambda ix: ix[0],
+    "last":           lambda ix: ix[-1],
+    "min(caps[:1])":  lambda ix: min(ix[:1], key=lambda p: p[1]),
+    "min(caps[-1:])": lambda ix: min(ix[-1:], key=lambda p: p[1]),
+    "min(caps[:2])":  lambda ix: min(ix[:2], key=lambda p: p[1]),
+    "min(caps[:3])":  lambda ix: min(ix[:3], key=lambda p: p[1]),
+    "min(caps[1:])":  lambda ix: min(ix[1:], key=lambda p: p[1]),
+    "min(caps[:-1])": lambda ix: min(ix[:-1], key=lambda p: p[1]),
+    "min(caps[::2])": lambda ix: min(ix[::2], key=lambda p: p[1]),
+    "mean":           lambda ix: (0, sum(c for _i, c in ix) / len(ix)),
+}
+_survivors = []
+print("    the family table - each candidate against what the SHIPPED "
+      "validator answered on the same roster:")
+for _name, _fn in _FAMILY.items():
+    _misses = []
+    for _opt, _n, _pos, _caps, (_ok, _unit, _cap) in _rows:
+        _pred = _pick(list(_caps), _fn)
+        _pred_refuses = _pred is not None and _pred[1] > 0 and _RATE >= _pred[1]
+        if _opt == "maintenance_charge_w":
+            _pred_refuses = (_pred is not None and _pred[1] > 0
+                             and _RATE > _pred[1])
+        _said = ((True, _pred[0], _pred[1]) if _pred_refuses
+                 else (False, None, None))
+        if _said != (_ok, _unit, _cap):
+            _misses.append(f"{_opt[:4]}/n{_n}p{_pos}")
+    if not _misses and _name != "min (SHIPPED)":
+        _survivors.append(_name)
+    print(f"      {_name:<16} {'AGREES EVERYWHERE (survivor)' if not _misses else 'differs on ' + str(len(_misses)) + ' run(s), first: ' + _misses[0]}")
+check("⚠️ AND THE FAMILY TABLE HAS NO SURVIVOR BUT THE SHIPPED REDUCTION. "
+      "Eleven candidates - max, first, last, the mean, and every positional "
+      "slice this card has actually been defeated by - are each computed "
+      "against what the real validator answered on the same roster, and every "
+      "one of them differs somewhere in the set. This is the check that would "
+      "have taken PR #48 red four rounds ago: a member no fixture kills is "
+      "PRINTED AS A NAMED SURVIVOR here instead of being absent from a list "
+      f"nobody re-derived. survivors: {_survivors}",
+      _survivors == [])
+
+# --- the ROSTER OF ONE, where every one of them is the same thing -----------
+_at_one = {n: _pick([_SMALL], f) for n, f in _FAMILY.items()}
+print(f"    at a ROSTER OF ONE the twelve candidates answer: "
+      f"{sorted({str(v) for v in _at_one.values()})}")
+check("...and at a ROSTER OF ONE every candidate that still has a member to "
+      "reduce answers THE SAME THING - which is exactly why no "
+      "single-inverter fixture in this repository could ever have caught any "
+      "of this, and why the shipped single-inverter refusals are what they "
+      "always were. The only two that do NOT coincide are `caps[1:]` and "
+      "`caps[:-1]`, which EMPTY the list and would raise: at N=1 they are "
+      "crashes rather than silent wrong answers, so they are not survivors "
+      "there either",
+      {v for v in _at_one.values() if v is not None} == {("unit-0", _SMALL)}
+      and {n for n, v in _at_one.items() if v is None}
+      == {"min(caps[1:])", "min(caps[:-1])"}
+      and _verdict(_roster_n([_SMALL], **dict(_QUIET, sign_disagree_w=_RATE)),
+                   "SAFETY-02 sign-disagreement")[0])
+
+# --- rule 5's SIBLING BRANCH, which was reading the FLAT max_w --------------
+# ⚠️ FOUND BY READING THE THREE BRANCHES SIDE BY SIDE, NOT BY THE STRUCTURAL
+# PIN - and said that way on purpose, because it is the shape of what the pin
+# cannot see. `unit_sign_disagree_w` is SAFETY-30's, not this card's. It was
+# graded against `_number(opts, "max_w", 2000.0)`, the FLAT option, three lines
+# under a branch that had already stopped doing that. Both directions were
+# measured before the fix and both are asserted here.
+_leftover = _roster_n([_SMALL, _SMALL],
+                      **dict(_QUIET, unit_sign_disagree_w=_RATE))
+_leftover["max_w"] = _BIG          # what a single-inverter install leaves behind
+_lv = _verdict(_leftover, "SAFETY-30 per-inverter sign-disagreement")
+print(f"    leftover flat max_w {_BIG:.0f} beside a roster of two "
+      f"{_SMALL:.0f} W units, unit_sign_disagree_w {_RATE:.0f} -> {_lv}")
+check("⚠️ A LEFTOVER FLAT `max_w` NO LONGER DISARMS SAFETY-30's PER-UNIT "
+      "WATCHDOG SILENTLY. Two 2000 W units beside a `max_w: 5000` left over "
+      "from a single-inverter install and `unit_sign_disagree_w: 3000` used to "
+      "produce ZERO problems - the site started, the status page said `armed`, "
+      "and the watchdog could not trip on EITHER unit. It is refused now, "
+      "against the weakest unit's own cap, and the message names that unit",
+      _lv == (True, "unit-0", _SMALL))
+check("...and the over-refusal in the OTHER direction is gone too: a roster of "
+      "two 5000 W units with NO flat `max_w` was refused at 3000 W against a "
+      "2000 W DEFAULT the operator never wrote, for a threshold both units can "
+      "reach. It starts",
+      _verdict(_roster_n([_BIG, _BIG],
+                         **dict(_QUIET, unit_sign_disagree_w=_RATE)),
+               "SAFETY-30 per-inverter sign-disagreement")
+      == (False, None, None))
+check("...and the flat single-inverter config it always refused is refused "
+      "IDENTICALLY - `unit_caps()` at a roster of one is `[max_w]`, so this is "
+      "the same comparison, the same boundary and the same sentence",
+      why(validate_startup_config(base(max_w=2000, unit_sign_disagree_w=2000),
+                                  -5000, 5000), "unit_sign_disagree_w")
+      and validate_startup_config(base(max_w=2000, unit_sign_disagree_w=1999),
+                                  -5000, 5000) == [])
+
+
 # --------------------------------------------------------------------------- #
 print()
 if fails:

goodwe_controller/test_watchdog.py

@@ -401,10 +401,19 @@ print("controller and the first thing an operator will try")
 
 
 class _MemStore:
-    """A Store that keeps its dict across "restarts", like the real file does."""
+    """A Store that keeps its dict across "restarts", like the real file does.
+
+    ⚠️ SEEDED FROM `store.DEFAULTS`, NOT FROM A LITERAL. `Store.load()` merges
+    that dict over whatever is on disk, so a fake that hand-listed three keys
+    answered `None` where the real store answers a default - and a check reading
+    "this latch was never written" then could not tell an absent key from a
+    False one. It is also a place to drift: SAFETY-31 and SAFETY-30 both added
+    flat pairs, and neither would have appeared here.
+    """
 
     def __init__(self):
-        self.data = {"phase": "idle", "sign_tripped": False, "sign_reason": ""}
+        from app.store import DEFAULTS as _SDEF
+        self.data = dict(_SDEF)
 
     def set(self, k, v):
         self.data[k] = v
@@ -789,9 +798,18 @@ print("   circulation the allocator NEVER COMMANDED")
 
 from app.inverter import Inverter  # noqa: E402
 
-# ⚠️ INVERTER B DOES NOT PHYSICALLY EXIST - it is ~10 days out and SPIKE-04
-# measures it. Its rating here is a HYPOTHESIS, named as one, exactly as
-# test_peakguard.py's helper names its own.
+# ⚠️ INVERTER B DOES NOT PHYSICALLY EXIST - it is ~8 days out (2026-09-04) and
+# the bench half of SPIKE-04 measures it.
+# ⚠️ SPIKE-04a (2026-09-04) ESTABLISHED B'S REAL RATING FROM ITS DATASHEET:
+# a GoodWe GW5048D-ES, serial `55048ESU22AW0992`, 4600 W on the battery path
+# (goodwe/NOTES.md §1.1b). THIS FIXTURE DELIBERATELY DOES NOT USE IT. 3000 W is
+# kept because the only property these checks consume is that B's number is
+# UNMISTAKABLY not A's, and the real pair is 5000/4600 - a 400 W gap, narrower
+# than `slew_w` and inside the range where an off-by-one-unit bug and ordinary
+# control arithmetic produce indistinguishable numbers. Collapsing the fixture
+# onto the real values would cost these checks their separating power to buy
+# realism they do not use. So: ILLUSTRATIVE, NOT B'S RATING, and not a
+# placeholder for it either - the placeholder era ended on 2026-09-04.
 B_RATING_HYPOTHESIS = 3000.0
 
 
@@ -1105,6 +1123,876 @@ check("⚠️ AND THE 'HAS A HUMAN SEEN THIS' FLAG IS CLEARED WITH IT. Left "
       _cc._circ_announced is False)
 
 
+
+print()
+print("=" * 72)
+print("SAFETY-30: at two inverters the sign check is TWO questions, not one")
+print("=" * 72)
+# ⚠️ READ THIS BEFORE ADDING A CHECK HERE. Half of this card was already built
+# and it is NOT the half below. DEV-02 shipped the PER-INSTANCE half: each
+# `Inverter` owns its own `SignWatchdog`, latched under its own namespaced store
+# keys, and a trip on B no longer latches A. test_inverter.py part 3 grades
+# that; nothing here re-tests it.
+#
+# What is graded here is THE COMPARISON ITSELF. Before this card every unit's
+# detector was handed `self.grid`. There is ONE P1, so at N>=2 that is a number
+# this controller put on ONE actuator, graded against a measurement that answers
+# for ALL of them - wrong in both directions: a FALSE trip when a sibling
+# accounts for the meter, and a MISSED one when a sibling MASKS a unit that
+# really is reversed. The fix is two comparisons, and they are two different
+# faults: each unit against ITS OWN measured power, and the SUM of the standing
+# commands against the meter.
+#
+# ⚠️ AND IT IS NOT SAFETY-31, WHICH IS THE SECTION ABOVE. `circulation.py`
+# compares the units AGAINST EACH OTHER and catches two inverters pushing
+# THROUGH each other, which nets to zero at the meter. This compares each unit
+# against ITS OWN COMMAND and catches a unit not doing what it was told. They
+# overlap in exactly one scenario - a unit reversing while its sibling pulls the
+# other way - and part 4 below asserts what each one concludes there, because
+# two guards that overlap without saying so is how a false-alarm generator gets
+# built.
+
+from app.control import Tuning as _Tuning  # noqa: E402
+from app.watchdog import (  # noqa: E402
+    AGAINST_METER, AGAINST_OWN_POWER, SIGN_OPTION_DEFAULTS, disagrees,
+)
+
+print()
+print("1. ONE predicate, TWO second terms - and the trip sentence says which")
+# ⚠️ THE ARITHMETIC IS UNCHANGED AND THAT IS THE POINT: `disagrees()` was not
+# split in two. What moved is a PARAMETER NAME - `grid_w` became `against_w` -
+# because what the command is opposed TO is now the caller's decision. A second
+# copy of a safety predicate is one copy plus a future disagreement about what
+# the check means (see `loop_is_following()`), so there is still exactly one.
+_scfg = SignWatchdogConfig(min_w=300.0, hold_s=180.0)
+_ucfg = SignWatchdogConfig(
+    min_w=float(SIGN_OPTION_DEFAULTS["unit_sign_disagree_w"]),
+    hold_s=float(SIGN_OPTION_DEFAULTS["unit_sign_disagree_s"]))
+check("the same call that means 'commanded -1000 W into a +1500 W import' "
+      "means 'commanded -1000 W while measuring +1500 W of discharge' when the "
+      "caller hands it a unit's own power - one function, two questions",
+      disagrees(-1000.0, 1500.0, True, _scfg) is True)
+check("...and every gate still binds on the second term whatever it is: a "
+      "250 W reading does not qualify against a 300 W gate",
+      disagrees(-1000.0, 250.0, True, _scfg) is False)
+check("...nor does a unit sitting at zero, which is a REFUSED command and not "
+      "a reversed one - the distinction the module docstring makes about a "
+      "battery declining a discharge, now read of the unit's own sensor",
+      disagrees(1500.0, 0.0, True, _ucfg) is False)
+
+_wd_site = SignWatchdog(_scfg)
+_wd_unit = SignWatchdog(_ucfg, against=AGAINST_OWN_POWER)
+for _t in range(0, 400, 5):
+    _wd_site.evaluate(-1000.0, 1500.0, True, T0 + timedelta(seconds=_t))
+    _wd_unit.evaluate(-1000.0, 1500.0, True, T0 + timedelta(seconds=_t))
+print(f"       site: {_wd_site.reason!r}")
+print(f"       unit: {_wd_unit.reason!r}")
+check("⚠️ THE SITE INSTANCE'S SENTENCE IS BYTE-IDENTICAL TO THE ONE THAT "
+      "SHIPPED. It is what an operator reads on the status page, what goes in "
+      "the action log and what is persisted to /data/state.json, so it is "
+      "asserted against the literal rather than against a format string",
+      _wd_site.reason == (
+          "SIGN_DISAGREEMENT: commanded -1000 W against a meter reading "
+          "+1500 W for 180 s continuously (limits: both over 300 W, held over "
+          "180 s)"))
+check("...and the per-unit instance names what it ACTUALLY compared instead. "
+      "'against a meter reading' there would be a false statement about which "
+      "instrument produced the evidence, on the one line somebody woken by "
+      "this alarm reads first",
+      AGAINST_OWN_POWER in _wd_unit.reason
+      and AGAINST_METER not in _wd_unit.reason
+      and "+1500 W" in _wd_unit.reason)
+# ⚠️ `arm()` IS CONFIGURATION AND MUST NOT BE STATE. `Controller.cycle()` calls
+# it on EVERY cycle, so a version that cleared `disagree_since` there would
+# restart the window every cycle, never accumulate a hold, and leave the check
+# permanently green with nothing saying so. Driven rather than argued: open a
+# window, re-arm mid-episode, and see the window still standing.
+_rearm = SignWatchdog(_ucfg, against=AGAINST_OWN_POWER)
+_rearm.evaluate(-1000.0, 1500.0, True, T0)
+_rearm.evaluate(-1000.0, 1500.0, True, T0 + timedelta(seconds=5))
+_rearm.arm(_ucfg, AGAINST_OWN_POWER)
+check("re-arming mid-episode leaves the window exactly where it was",
+      _rearm.disagree_since == T0)
+for _t in range(10, 200, 5):
+    _rearm.arm(_ucfg, AGAINST_OWN_POWER)
+    _rearm.evaluate(-1000.0, 1500.0, True, T0 + timedelta(seconds=_t))
+check("...so a detector re-armed on every one of 38 cycles still trips at its "
+      "85 s hold, which is what says `arm()` did not quietly disarm it",
+      _rearm.tripped)
+
+print()
+print("2. CFG-03: config.yaml is the source of truth and there is ONE copy")
+_sign_from_yaml = {k: int(v) for k, v in
+                   _re.findall(r"^  ((?:unit|site)_sign_disagree_\w+):\s*(\d+)\s*$",
+                               _cfg_opts, _re.M)}
+print(f"       config.yaml options: {_sign_from_yaml}")
+print(f"       SIGN_OPTION_DEFAULTS: {SIGN_OPTION_DEFAULTS}")
+check("all four new options are declared in config.yaml's options: block and "
+      "the in-code mirror agrees key for key and value for value - the "
+      "Dockerfile does not put config.yaml in the image, so the running add-on "
+      "cannot read it and this is where the two can drift",
+      _sign_from_yaml == SIGN_OPTION_DEFAULTS and len(_sign_from_yaml) == 4)
+_sign_bounds = {m[0]: (int(m[1]), int(m[2])) for m in
+                _re.findall(r"^  ((?:unit|site)_sign_disagree_\w+):"
+                            r"\s*int\((\d+),(\d+)\)\s*$",
+                            _cfg_schema, _re.M)}
+print(f"       schema bounds: {_sign_bounds}")
+check("...and ALL FOUR have a schema range, so the Supervisor refuses an "
+      "out-of-range value rather than the add-on discovering it at runtime",
+      set(_sign_bounds) == set(SIGN_OPTION_DEFAULTS))
+check("the shipped defaults sit inside their own schema ranges",
+      all(_sign_bounds[k][0] <= v <= _sign_bounds[k][1]
+          for k, v in SIGN_OPTION_DEFAULTS.items()))
+check("⚠️ THE THRESHOLD IS `saturation_w / 2`, WHICH IS WHERE THE CONTROL LAW "
+      "ITSELF ALREADY DRAWS THE LINE. This check gates on BOTH sides, and the "
+      "two are OPPOSED, so the smallest command-versus-power gap it can ever "
+      "see is 2 x min_w. At 250 W that is exactly the 500 W at which "
+      "`compute()`'s own anti-windup test concludes the inverter is not "
+      "following and freezes the integrator - so this watchdog cannot fire on "
+      "a divergence the shipped loop calls ordinary tracking, and the two "
+      "guards meet at a number instead of overlapping in a band nobody drew",
+      SIGN_OPTION_DEFAULTS["unit_sign_disagree_w"] * 2
+      == _Tuning().saturation_w)
+check("...and the window clears its three bounds: above the firmware's own "
+      "30 s link veto (`inverter.LINK_STALE_S` - inside it, a unit not "
+      "executing its share is a LINK fault the add-on already answers by "
+      "dropping it from the split), and UNDER HALF the 180 s the site check "
+      "takes, because the per-unit conclusion needs nothing to net out before "
+      "it is complete",
+      SIGN_OPTION_DEFAULTS["unit_sign_disagree_s"] > LINK_STALE_S
+      and SIGN_OPTION_DEFAULTS["unit_sign_disagree_s"] < 180 / 2)
+check("⚠️ AND NEITHER IS A ROUND NUMBER. 250 W is read off `saturation_w`; "
+      "85 s is 8.5x the worst legitimate opposition 1890 two-inverter runs "
+      "produce (10.0 s) and 2.8x the firmware veto. 60, 75, 90, 100, 120, 180 "
+      "and 300 are all values a guess would have landed on",
+      SIGN_OPTION_DEFAULTS["unit_sign_disagree_s"]
+      not in (30, 60, 75, 90, 100, 120, 180, 300))
+
+# ⚠️ AND THE SITE PAIR IS TWO DECISIONS, NOT ONE, WHICH IS WHY BOTH ARE PINNED.
+# The THRESHOLD is carried from the shipped pair because its argument is about
+# the meter's noise floor and the deadband - a property of ONE P1 and of
+# `Tuning.deadband_w`, neither of which the roster touches. The WINDOW is NOT
+# carried, because the transient it rejects is the loop crossing zero across a
+# rail that IS the roster sum: measured at 110 s with one inverter and 215 s
+# with two, on the same stimuli, against a 180 s hold. That is a carried number
+# justified and a carried number refused, in the same pair.
+check("the SITE threshold is the shipped one, carried - the noise argument "
+      "behind 300 W is about the meter and the deadband, and a second inverter "
+      "changes neither",
+      SIGN_OPTION_DEFAULTS["site_sign_disagree_w"] == 300)
+check("⚠️ ...AND THE SITE WINDOW IS LONGER THAN THE SHIPPED 180 s, WHICH IS THE "
+      "MEASUREMENT REFUSING TO BE CARRIED. At two inverters the worst "
+      "legitimate opposition the rig produces is 215.0 s - reachable at "
+      "`gain: 0.05`, `slew_w: 50` and a 2.2 kW load reversal, all inside what "
+      "config.yaml accepts - so 180 s would have shipped a false trip on the "
+      "check with the best record in the safety table. 325 s is 1.51x that",
+      SIGN_OPTION_DEFAULTS["site_sign_disagree_s"] > 215.0
+      and SIGN_OPTION_DEFAULTS["site_sign_disagree_s"] < 215.0 * 2)
+check("...and it is still inside the 900 s quarter that bills the capacity "
+      "tariff, and inside every episode the 2026-08-20 fault day contains - "
+      "its four clusters run 16, 30, 77 and 77 minutes",
+      SIGN_OPTION_DEFAULTS["site_sign_disagree_s"] < 900
+      and SIGN_OPTION_DEFAULTS["site_sign_disagree_s"] < 16 * 60)
+check("...and it is deliberately not 300 or 360, the two round numbers next to "
+      "it",
+      SIGN_OPTION_DEFAULTS["site_sign_disagree_s"] not in (300, 360, 600, 900))
+
+print()
+print("3. AC 1: ONE unit drives the wrong way, and ONLY that unit trips")
+# ⚠️ THE PLANT IS MODELLED AND SAYS SO. `_SignHass` is two inverters that obey
+# INSTANTLY - no ramp, no lag - and one meter that is `load - sum(measured)`.
+# That is deliberately not a claim about transients: the worst legitimate
+# opposition a real ramp and lag can produce is measured on admin/goodwe's
+# two-master rig, over 1890 closed-loop runs, and recorded in
+# DECISIONS/SAFETY-30-per-unit-sign.md. What this fixture is for is the
+# ATTRIBUTION - which detector concludes what, from which instrument - and for
+# that an instant plant is the honest model, because it removes the one
+# mechanism that could make a unit look reversed without being reversed.
+
+
+class _SignHass(_Hass):
+    """Two inverters, one meter, and a chosen subset that does the OPPOSITE.
+
+    `reversed_units` are roster indices whose measured power is the NEGATION of
+    what this controller told them - HW-01's wedged power stage, an inverted
+    driver, a unit executing a stale share backwards. Everything else obeys.
+    """
+
+    def __init__(self, load_w=2000.0, reversed_units=()):
+        super().__init__()
+        self.load_w = float(load_w)
+        self.reversed_units = set(reversed_units)
+        self.ctrl = None
+
+    def measured(self):
+        if self.ctrl is None:
+            return [0.0, 0.0]
+        out = []
+        for i, inv in enumerate(self.ctrl.inverters):
+            w = inv.last_write if inv.last_write is not None else inv.target
+            w = float(w or 0.0)
+            out.append(-w if i in self.reversed_units else w)
+        return out
+
+    async def number(self, entity, invert=False):
+        m = self.measured() + [0.0, 0.0]
+        return {"sensor.grid": self.load_w - sum(m[:2]),
+                "sensor.soc": 50.0, "sensor.soc_b": 50.0,
+                "sensor.batt": m[0], "sensor.batt_b": m[1]}.get(entity)
+
+
+def _sign_site(hass, store=None, **over):
+    """A two-inverter Controller, with SAFETY-31 deliberately given a long window.
+
+    ⚠️ THE LONG `circulation_hold_s` IS ISOLATION, NOT A CLAIM THAT SAFETY-31 IS
+    SILENT HERE. A unit reversing while its sibling pulls the other way IS
+    circulating power and SAFETY-31 sees it - part 4 asserts exactly that, and
+    what each guard concludes about it. What must not happen is the SITE being
+    stopped by circulation while this section is trying to grade a PER-UNIT
+    attribution, because then every "only B is parked" check below would be
+    measuring the wrong guard.
+
+    ⚠️ THE THREE THRESHOLD PAIRS ARE PAIRWISE DISTINCT IN BOTH COORDINATES,
+    AND THAT IS LOAD-BEARING RATHER THAN COSMETIC. This fixture is the only
+    place in the suite where `_arm_sign_watchdogs()` runs at a roster of two, so
+    it is the only place that can see WHICH pair reaches WHICH detector. The
+    first version of it set every hold to 1 s and two of the three magnitude
+    gates to 300 W; a detector armed with the shipped pair, or with the site
+    pair, or with the right magnitude and the wrong hold, then compared EQUAL
+    to the pair it was supposed to have, and the wiring check below would have
+    passed on three of those four arrangements. That is MUTATION.md's third
+    failure mode written into a fixture - a check comparing two things that
+    happen to be equal cannot see the difference it names - and the
+    precondition beside the check is what stops a later edit collapsing them
+    again without a red line.
+
+    Anyone changing a hold here: the PER-UNIT hold is what part 3's trip timing
+    is read against, and the SITE hold must still expire inside `_drive()`'s
+    window (14 cycles x 0.2 s), which part 5 depends on.
+    """
+    opts = {"setpoint_entity": "number.sp", "meter_entity": "sensor.grid",
+            "soc_entity": "sensor.soc", "batt_entity": "sensor.batt",
+            "meter_source": "off", "max_w": 2000, "gain": 0.6,
+            "sign_disagree_w": 300, "sign_disagree_s": 3,
+            "unit_sign_disagree_w": 250, "unit_sign_disagree_s": 1,
+            "site_sign_disagree_w": 320, "site_sign_disagree_s": 2,
+            "circulation_hold_s": 3600, "auto_start": True}
+    opts.update(over)
+    c = Controller(opts, hass, store if store is not None else _MemStore(),
+                   _P(), stop=asyncio.Event())
+    b = Inverter("inverter-b", setpoint_entity="number.sp_b",
+                 soc_entity="sensor.soc_b", batt_entity="sensor.batt_b",
+                 max_w=B_RATING_HYPOTHESIS, store=c.store)
+    b.dev_min, b.dev_max = -B_RATING_HYPOTHESIS, B_RATING_HYPOTHESIS
+    c.inverters.append(b)
+    c.auto = True
+    for inv in c.inverters:
+        if inv.dev_min is None:
+            inv.dev_min, inv.dev_max = -5000.0, 5000.0
+    hass.ctrl = c
+    return c
+
+
+async def _drive(c, n=14, dt=0.2, watch=None):
+    seen = []
+    for _ in range(n):
+        await c.read_inputs()
+        await c.cycle()
+        if watch is not None:
+            seen.append(watch(c))
+        await asyncio.sleep(dt)
+    return seen
+
+
+async def _one_reversed():
+    hass = _SignHass(load_w=2000.0, reversed_units={1})
+    c = _sign_site(hass)
+    seen = await _drive(c, watch=lambda c: dict(
+        grid=c.grid,
+        stood=[i.last_write if i.last_write is not None else i.target
+               for i in c.inverters],
+        meas=[i.batt for i in c.inverters],
+        a=c.inverters[0].sign_wd.tripped, b=c.inverters[1].sign_wd.tripped,
+        site=c.site_sign_wd.tripped))
+    return c, hass, seen
+
+
+_c30, _h30, _s30 = asyncio.run(_one_reversed())
+_a30, _b30 = _c30.inverters
+
+# ⚠️ THE WIRING PIN FOR AC 5, AND IT IS A DIFFERENT CLAIM FROM EVERY
+# BEHAVIOURAL CHECK IN THIS SECTION. The rest of part 3 grades what the
+# detectors CONCLUDED; this grades WHICH THRESHOLDS THEY WERE HANDED, and the
+# gap between the two was live. Replacing
+# `arm(self.sign_cfg if one else self.unit_sign_cfg, ...)` with
+# `arm(self.sign_cfg, ...)` in `Controller._arm_sign_watchdogs()` throws the
+# re-derived per-unit pair away and arms every unit with SAFETY-02's
+# meter-facing one - and left ALL 19 FILES GREEN. Nothing at a roster of two
+# read `inv.sign_wd.cfg` at all; the one place that does is part 7's N=1 arm,
+# which that edit PRESERVES.
+#
+# WHY THE TRIPS BELOW CANNOT CATCH IT, said here because "the suite is green"
+# and "nothing was wrong" are the two readings this comment exists to separate.
+# B is reversed at ~2000 W - far over 250 W and far over 300 W - and every hold
+# in this fixture is short enough to expire inside `_drive()`. A fault that is
+# big enough and long enough to trip under BOTH pairs is evidence about
+# neither. Only an equality on the pair itself can tell them apart.
+#
+# WHAT THE SURVIVOR COSTS IN THE FIELD, at config.yaml's own defaults: the
+# per-unit check runs at `sign_disagree_s` 180 s instead of the re-derived 85 s
+# - 2.1x slower to park a reversed inverter - and its gate moves from 250 W to
+# 300 W, so a unit reversing a share between the two becomes invisible, while
+# watchdog.SIGN_OPTION_DEFAULTS, config.yaml and DECISIONS section 3 all go on
+# describing a pair the running code does not use.
+_UNIT_PAIR = (250.0, 1.0)      # `unit_sign_disagree_w` / `_s`  - one per unit
+_SITE_PAIR = (320.0, 2.0)      # `site_sign_disagree_w` / `_s`  - one per site
+_SHIPPED_PAIR = (300.0, 3.0)   # `sign_disagree_w` / `_s` - N=1 only
+_armed = {inv.name: (inv.sign_wd.cfg.min_w, inv.sign_wd.cfg.hold_s)
+          for inv in _c30.inverters}
+_armed_site = (_c30.site_sign_wd.cfg.min_w, _c30.site_sign_wd.cfg.hold_s)
+print(f"       armed per unit: {_armed}")
+print(f"       armed site:     {_armed_site}    "
+      f"shipped pair, wanted by NOBODY here: {_SHIPPED_PAIR}")
+check("PRECONDITION: the three pairs are pairwise distinct IN BOTH "
+      "COORDINATES, so an equality below can fail on the magnitude alone, on "
+      "the hold alone, or on both. This fixture originally gave all three the "
+      "same 1 s hold and two of them the same 300 W gate, and a mis-wiring "
+      "that delivered the right magnitude with the wrong hold would then have "
+      "compared EQUAL to the pair it was supposed to have",
+      len({_UNIT_PAIR, _SITE_PAIR, _SHIPPED_PAIR}) == 3
+      and len({p[0] for p in (_UNIT_PAIR, _SITE_PAIR, _SHIPPED_PAIR)}) == 3
+      and len({p[1] for p in (_UNIT_PAIR, _SITE_PAIR, _SHIPPED_PAIR)}) == 3)
+check("PRECONDITION: and all three are pairs this Controller actually holds, "
+      "so the literals below are the fixture's own numbers and not three "
+      "constants that agree with nothing. In particular the shipped pair is "
+      "still constructed and still a reachable mis-wiring target",
+      (_c30.unit_sign_cfg.min_w, _c30.unit_sign_cfg.hold_s) == _UNIT_PAIR
+      and (_c30.site_sign_cfg.min_w, _c30.site_sign_cfg.hold_s) == _SITE_PAIR
+      and (_c30.sign_cfg.min_w, _c30.sign_cfg.hold_s) == _SHIPPED_PAIR)
+check("⚠️ AC 5 WIRING. EVERY per-unit detector on this roster of two is armed "
+      "with the PER-UNIT pair - 250 W / 1 s here, `unit_sign_disagree_w` and "
+      "`unit_sign_disagree_s` in the field. Graded against LITERALS rather "
+      "than against `_c30.unit_sign_cfg`, which cannot tell 'the per-unit "
+      "pair' from 'whatever this Controller was handed'. This is the line that "
+      "goes red when `_arm_sign_watchdogs()` stops choosing between the two "
+      "pairs and hands every unit the site-facing one",
+      len(_armed) == 2 and all(p == _UNIT_PAIR for p in _armed.values()))
+check("...and the SITE detector is armed with the SITE pair - 320 W / 2 s "
+      "here - which differs from the per-unit pair AND from the shipped pair "
+      "in both coordinates, so this equality is not carried by a number the "
+      "fixture happens to hold twice",
+      _armed_site == _SITE_PAIR)
+check("...and each detector is pointed at its own INSTRUMENT as well as at its "
+      "own thresholds: the units at their own 0x0518 power, the site at the "
+      "meter. They are two arguments of one call and a pin on one is not a pin "
+      "on the other",
+      all(inv.sign_wd.against == AGAINST_OWN_POWER for inv in _c30.inverters)
+      and _c30.site_sign_wd.against == AGAINST_METER)
+
+_mid = _s30[3]
+print(f"       cycle 3: stood {[round(x) for x in _mid['stood']]}  "
+      f"measured {[round(x) for x in _mid['meas']]}  "
+      f"meter {_mid['grid']:+.0f} W")
+check("PRECONDITION: the stimulus is not vacuous - inverter B really is doing "
+      "the OPPOSITE of what it was told, and inverter A really is obeying. "
+      "Without this line every claim below could be true of a fixture in which "
+      "nothing was reversed",
+      _mid["stood"][1] * _mid["meas"][1] < 0
+      and abs(_mid["meas"][1]) > 250
+      and _mid["stood"][0] * _mid["meas"][0] > 0)
+check("⚠️ AC 1. THE REVERSED UNIT TRIPS AND ITS SIBLING DOES NOT. The evidence "
+      "is B's own 0x0518 power against B's own share, so it is a fact about B "
+      "and nothing about A - which is what makes parking one of them the right "
+      "remedy rather than an attribution nobody measured",
+      _b30.sign_wd.tripped and not _a30.sign_wd.tripped)
+check("...and the reason names the instrument, so the operator is sent to the "
+      "right one", AGAINST_OWN_POWER in _b30.sign_wd.reason)
+print(f"       {_b30.sign_wd.reason!r}")
+# ⚠️ READ OFF THE ALLOCATION RECORD, NOT OFF `Controller.reason`. At a roster
+# of two `reason` is the SITE resolution's line - "loop -> 2000 W" - because
+# B's failsafe is a UNIT claim resolved on B's own actuator and never competes
+# for the site winner. That is the DEV-02 split working, and a check that
+# grepped the site line for "failsafe" would go green only while the roster was
+# one. The per-unit record is where the pin is stated, and it names its source.
+_bplan = [pl for pl in (_c30.allocation.plans if _c30.allocation else ())
+          if pl.unit == "inverter-b"]
+print(f"       allocation: {_c30.allocation.explain()}")
+print(f"       inverter-b plan: {_bplan[0].why if _bplan else '(none)'}")
+check("...B is commanded 0 W by its OWN `failsafe` claim, named in the "
+      "allocation record, rather than by some other guard happening to stop it",
+      _b30.target == 0.0 and _bplan and "failsafe" in _bplan[0].why)
+check("...and A IS STILL COMMANDING. One inverter in trouble stops that "
+      "inverter; the site goes on running on the one that works, which is what "
+      "a UNIT-scoped claim buys and what a site-wide stop would have thrown "
+      "away", _a30.target != 0.0)
+check("...the latch is under B's OWN namespaced key, and A's is untouched",
+      _c30.store.data.get(unit_key("inverter-b", "sign_tripped")) is True
+      and _c30.store.data.get(TRIPPED_KEY) is False)
+
+print()
+print("4. AC 2: and it is caught EVEN THOUGH THE METER SAYS EVERYTHING IS FINE")
+# ⚠️ THIS IS THE CHECK THE WHOLE CARD IS FOR, so it is built the way MUTATION.md
+# says to build one: the claim is "the new comparison sees what the old one
+# structurally cannot", and the OLD comparison is RUN, on the same samples, and
+# its blindness is asserted as its own precondition and PRINTED - not argued.
+_old_view = [disagrees(s["stood"][1], s["grid"], True, _scfg) for s in _s30]
+_site_view = [disagrees(sum(s["stood"]), s["grid"], True, _scfg) for s in _s30]
+print(f"       the SHIPPED comparison (B's command vs the meter) was true on "
+      f"{sum(_old_view)} of {len(_old_view)} cycles")
+print(f"       the SITE comparison (sum of commands vs the meter) on "
+      f"{sum(_site_view)} of {len(_site_view)}")
+check("⚠️ THE COMPARISON THAT SHIPPED IS FALSE ON EVERY SINGLE CYCLE HERE. B is "
+      "commanded to discharge and the house is importing, so B's command and "
+      "the meter AGREE all the way through - the meter cannot tell that the "
+      "watts arriving are inverter A's while B does the opposite. This is not "
+      "a threshold set too high: the summation destroys the evidence and no "
+      "tuning recovers it",
+      not any(_old_view))
+check("...and the SITE comparison is blind for the same reason, which is why "
+      "it is kept as a SEPARATE question rather than offered as the fix",
+      not any(_site_view))
+check("...while the shipped SITE watchdog stayed correctly silent, so nothing "
+      "here is a site-level trip wearing a per-unit label",
+      not _c30.site_sign_wd.tripped
+      and _c30.store.data.get("site_sign_tripped") is not True)
+check("⚠️ AND SAFETY-31 SEES SOMETHING TOO, WHICH IS NOT A DUPLICATE AND IS "
+      "SAID HERE RATHER THAN LEFT FOR SOMEBODY TO DISCOVER. A unit reversing "
+      "while its sibling pulls the other way IS power going round a loop, so "
+      "`circulating_w()` is over its threshold on this fixture - it is only "
+      "the long window this fixture gives it that keeps it from stopping the "
+      "site. The two guards CONCLUDE different things: circulating power is "
+      "defined over the roster and names no offender, while this one names B "
+      "and parks B. Neither is the other with a different threshold",
+      max(circulating_w(s["meas"]) for s in _s30) > _c30.circ_wd.cfg.min_w
+      and not _c30.circ_wd.tripped
+      and _c30.site_circulating_w == 0.0)
+
+print(f"       circulating power over the episode: "
+      f"{[round(circulating_w(s['meas'])) for s in _s30]}")
+check("...and it falls back to ZERO the moment B is parked, which is the "
+      "remedy working rather than the sensor going quiet: a unit held at 0 W "
+      "cannot be half of a loop. It is asserted on the EPISODE rather than on "
+      "the end state for exactly that reason",
+      _s30[-1]["meas"][1] == 0.0)
+
+print()
+print("5. AC 3: the SITE question survives, separately, and catches whole-site")
+print("   reversal - which no per-unit check can see")
+# ⚠️ THE FAULT HERE IS THE ONE THE 2026-08-20 CAPTURE CONTAINS, and it is
+# INVISIBLE to every per-unit comparison by construction: both inverters are
+# doing EXACTLY what they were told. What is wrong is the number they were told.
+# A negative gain is the minimal faithful model of it - an inverted polarity, a
+# wound-up integrator, a strategy on the wrong side of the meter - and it is not
+# a reachable configuration (`gain` is float(0.05,1.0)), which is the point: it
+# stands in for a code regression, which is the case a check can hold the line on.
+
+
+async def _site_reversed():
+    hass = _SignHass(load_w=2000.0, reversed_units=())
+    c = _sign_site(hass, gain=-0.6)
+    seen = await _drive(c, watch=lambda c: dict(
+        grid=c.grid, site=c.site_sign_wd.tripped,
+        stood=[i.last_write if i.last_write is not None else i.target
+               for i in c.inverters],
+        meas=[i.batt for i in c.inverters]))
+    return c, hass, seen
+
+
+_c31, _h31, _s31 = asyncio.run(_site_reversed())
+# ⚠️ THE LAST SAMPLE BEFORE THE TRIP, NOT A FIXED INDEX. After the trip every
+# command is 0 W by design, so a fixed index reads the REMEDY rather than the
+# fault and the precondition below would be asserting that a stopped site is
+# stopped. This is `Run.p2p_between`'s lesson from the rig, in a unit test: name
+# the window rather than inferring it.
+_pre31 = [x for x in _s31 if not x["site"]]
+_mid31 = _pre31[-1]
+print(f"       last cycle before the trip: stood "
+      f"{[round(x) for x in _mid31['stood']]}  "
+      f"measured {[round(x) for x in _mid31['meas']]}  "
+      f"meter {_mid31['grid']:+.0f} W")
+check("PRECONDITION: EVERY inverter is obeying - each unit's measured power "
+      "agrees in sign with its own share - and the site is still charging into "
+      "an importing house. Without this the site trip below could be a "
+      "per-unit fault in disguise",
+      all(s * m > 0 for s, m in zip(_mid31["stood"], _mid31["meas"]))
+      and sum(_mid31["stood"]) * _mid31["grid"] < 0)
+check("⚠️ AC 3. THE SITE WATCHDOG TRIPS AND NOT ONE PER-UNIT WATCHDOG DOES. "
+      "That asymmetry is the card: the fault is in the number the site was "
+      "asking for, which no single inverter owns, and every unit is innocent "
+      "of it",
+      _c31.site_sign_wd.tripped
+      and not any(i.sign_wd.tripped for i in _c31.inverters))
+print(f"       {_c31.site_sign_wd.reason!r}")
+check("...and it stops the WHOLE site, through the `sign_site` claim by name "
+      "rather than by some other guard happening to stop it",
+      _c31.target == 0.0 and "sign_site" in _c31.reason
+      and all(i.target == 0.0 for i in _c31.inverters))
+check("...the latch is FLAT - `site_sign_tripped`, beside SAFETY-31's pair - "
+      "and NOT mirrored onto the pre-DEV-02 `sign_tripped`, which is the "
+      "rollback view of a UNIT latch. A downgraded add-on drives exactly one "
+      "actuator; handing it a latch taken on evidence about a roster it cannot "
+      "have would be a trip it can never explain",
+      _c31.store.data.get("site_sign_tripped") is True
+      and "SIGN_DISAGREEMENT" in str(_c31.store.data.get("site_sign_reason"))
+      and _c31.store.data.get("sign_tripped") is not True)
+check("the status page carries a SITE row of its own, and the banner says the "
+      "site's TOTAL command opposed the meter rather than naming an inverter - "
+      "naming one would send an operator to swap the wrong box",
+      any("SITE SIGN-DISAGREEMENT FAILSAFE" in ck["text"]
+          for ck in _c31.checks())
+      and "total command" in _c31.status_payload()["banner"])
+_pub30 = {}
+_c31.mqtt = type("P", (), {"enabled": True,
+                           "publish": lambda _s, v: _pub30.update(v),
+                           "close": lambda _s: None})()
+_c31.publish()
+check("...and the controller-status sensor reports the failsafe state, so a "
+      "site parked at 0 W cannot publish 'running'",
+      _pub30.get("status") == "failsafe")
+
+# The latch, and the one way out of it.
+_was31 = _c31.site_sign_wd.tripped
+asyncio.run(_c31.handle_action("auto_toggle"))     # stop
+check("pressing STOP clears nothing - the condition always goes away on its "
+      "own, because the trip commands 0 W and a site at zero cannot oppose "
+      "anything", _c31.site_sign_wd.tripped and _was31)
+asyncio.run(_c31.handle_action("auto_toggle"))     # the deliberate start
+check("pressing START clears the latch in memory and on disk together, which "
+      "is the same deliberate human act that clears every other latch here",
+      not _c31.site_sign_wd.tripped
+      and _c31.store.data.get("site_sign_tripped") is False
+      and _c31.store.data.get("site_sign_reason") == "")
+
+
+async def _site_restart(store):
+    hass = _SignHass(load_w=2000.0)
+    c = _sign_site(hass, store=store, gain=-0.6)
+    await _drive(c, n=3, dt=0.05)
+    return c
+
+
+_store31 = _MemStore()
+_store31.data["site_sign_tripped"] = True
+_store31.data["site_sign_reason"] = "SIGN_DISAGREEMENT: (from before the restart)"
+_c31r = asyncio.run(_site_restart(_store31))
+check("⚠️ AND IT SURVIVES A RESTART, which is the only thing that makes it a "
+      "latch rather than a pause. 'Restart the add-on' is the first thing an "
+      "operator staring at a FAILSAFE banner tries, and a site that came back "
+      "armed would drive straight into the fault it had just stopped for",
+      _c31r.site_sign_wd.tripped and _c31r.target == 0.0
+      and "sign_site" in _c31r.reason)
+
+print()
+print("6. AC 4: two inverters commanded OPPOSITE SIGNS by a legitimate")
+print("   allocation, and nothing trips")
+# ⚠️ THIS IS THE FALSE TRIP THE CARD OPENS WITH, AND IT IS DRIVEN THROUGH THE
+# SHIPPED PREDICATE RATHER THAN DESCRIBED. Inverter A charging 1500 W while
+# inverter B discharges 1500 W into a 1500 W import is an ordinary two-inverter
+# arrangement - a per-unit maintenance charge on A while the loop drives B, or
+# SAFETY-25's floor pinning one unit. Nothing is faulty. Before this card the
+# meter convicted A on it.
+_A_CMD, _B_CMD, _METER = -1500.0, 1500.0, 1500.0
+_wd_a = SignWatchdog(_ucfg, against=AGAINST_OWN_POWER)
+_wd_b = SignWatchdog(_ucfg, against=AGAINST_OWN_POWER)
+_sitecfg = SignWatchdogConfig(
+    min_w=float(SIGN_OPTION_DEFAULTS["site_sign_disagree_w"]),
+    hold_s=float(SIGN_OPTION_DEFAULTS["site_sign_disagree_s"]))
+_wd_s = SignWatchdog(_sitecfg)
+for _t in range(0, 1200, 5):
+    _wd_a.evaluate(_A_CMD, _A_CMD, True, T0 + timedelta(seconds=_t))
+    _wd_b.evaluate(_B_CMD, _B_CMD, True, T0 + timedelta(seconds=_t))
+    _wd_s.evaluate(_A_CMD + _B_CMD, _METER, True, T0 + timedelta(seconds=_t))
+print(f"       A commanded {_A_CMD:+.0f} W and measured {_A_CMD:+.0f} W; "
+      f"B {_B_CMD:+.0f} / {_B_CMD:+.0f}; meter {_METER:+.0f} W")
+check("PRECONDITION: THE COMPARISON THAT SHIPPED CONVICTS INVERTER A HERE, on "
+      "every one of these samples. A is commanded to CHARGE while the meter "
+      "reads IMPORT - opposed, both far over 300 W, and it would have held for "
+      "20 minutes. That is the false trip, and it is asserted rather than "
+      "asserted about",
+      disagrees(_A_CMD, _METER, True, _scfg) is True)
+check("⚠️ AC 4. NEITHER UNIT TRIPS UNDER THE NEW COMPARISON, over 20 minutes at "
+      "an 85 s hold. Each one is doing exactly what it was told, and its own "
+      "power sensor is the instrument that can say so",
+      not _wd_a.tripped and not _wd_b.tripped)
+check("...and the SITE watchdog does not trip either, because the shares SUM "
+      "to zero and a 0 W total is not a wrong-direction command - the "
+      "magnitude gate that has always been there, now applied to the quantity "
+      "it belongs on", not _wd_s.tripped and _wd_s.disagree_since is None)
+# ⚠️ AND THE COUNTERFACTUAL, BECAUSE "nothing tripped" IS ALSO WHAT A DISARMED
+# GUARD REPORTS. Same site, same meter, but the shares no longer cancel: a NET
+# -1600 W of charge into a +1500 W import is the site genuinely driving the
+# wrong way, and it must trip. Without this line the check above passes
+# identically against a site watchdog that was never wired to anything.
+_wd_s_net = SignWatchdog(_sitecfg)
+for _t in range(0, 600, 5):
+    _wd_s_net.evaluate(-1600.0, _METER, True, T0 + timedelta(seconds=_t))
+check("...and the site question is not thereby disarmed: the same site, "
+      "commanded a NET -1600 W into the same import, DOES trip",
+      _wd_s_net.tripped)
+
+print()
+print("7. AC 8: at ONE inverter this card changes NOTHING, and 'nothing' is")
+print("   asserted rather than assumed")
+
+
+async def _n1_sign():
+    hass = _SignHass(load_w=2000.0)
+    opts = {"setpoint_entity": "number.sp", "meter_entity": "sensor.grid",
+            "soc_entity": "sensor.soc", "batt_entity": "sensor.batt",
+            "meter_source": "off", "max_w": 2000, "gain": -0.6,
+            "sign_disagree_w": 300, "sign_disagree_s": 1,
+            "auto_start": True}
+    c = Controller(opts, hass, _MemStore(), _P(), stop=asyncio.Event())
+    c.auto = True
+    c.dev_min, c.dev_max = -5000.0, 5000.0
+    hass.ctrl = c
+    # ⚠️ A CALL COUNTER, NOT AN OBSERVATION OF THE DETECTOR'S STATE, AND A
+    # MUTATION IS WHY. The first version of the check below read
+    # `site_sign_wd.disagree_since is None` - and arming the site detector at
+    # N=1 as well left this suite GREEN, because the unit detector trips first
+    # at its 1 s hold, parks the inverter, and the site's standing command is
+    # then 0 W: the site predicate is false, so its window is never opened and
+    # "never evaluated" reads identically to "evaluated and found nothing". That
+    # is MUTATION.md's third failure mode - the assertion ran, the sentence was
+    # true, and the data could not reach the failure it names. A counter can
+    # only be zero for the one reason.
+    _n1_calls = []
+    _orig_eval = c.site_sign_wd.evaluate
+
+    def _spy(*a, **kw):
+        _n1_calls.append(a)
+        return _orig_eval(*a, **kw)
+
+    c.site_sign_wd.evaluate = _spy
+    await _drive(c, n=12, dt=0.2)
+    return c, _n1_calls
+
+
+_c1s, _n1_site_calls = asyncio.run(_n1_sign())
+_inv1 = _c1s.inverters[0]
+print(f"       roster={_c1s.units}  armed against {_inv1.sign_wd.against!r} "
+      f"at {_inv1.sign_wd.cfg.min_w:.0f} W / {_inv1.sign_wd.cfg.hold_s:.0f} s")
+check("the single unit is graded against THE METER, at the SHIPPED thresholds "
+      "- §6.2's identity, and the whole compatibility promise of this card. "
+      "The unit's own power is not consulted, and `unit_sign_disagree_*` does "
+      "nothing on a one-inverter install",
+      _inv1.sign_wd.against == AGAINST_METER
+      and (_inv1.sign_wd.cfg.min_w, _inv1.sign_wd.cfg.hold_s) == (300.0, 1.0))
+check("...it trips, on the same evidence and through the same UNIT `failsafe` "
+      "claim as before, commanding 0 W",
+      _inv1.sign_wd.tripped and _c1s.target == 0.0
+      and "failsafe" in _c1s.reason and "sign_site" not in _c1s.reason)
+check("...the reason sentence still says 'a meter reading', word for word - it "
+      "is what an operator reads, what the action log carries and what is "
+      "persisted, so it is graded on the literal",
+      "against a meter reading" in _inv1.sign_wd.reason)
+check("...the latch is under the same namespaced key with the same flat "
+      "rollback mirror, and SAFETY-30's site pair is untouched at its default. "
+      "A one-inverter install's state file is byte-identical to what it was",
+      _c1s.store.data.get(TRIPPED_KEY) is True
+      and _c1s.store.data.get("sign_tripped") is True
+      and _c1s.store.data.get("site_sign_tripped") is False
+      and _c1s.store.data.get("site_sign_reason") == "")
+print(f"       site detector evaluated {len(_n1_site_calls)} times over "
+      f"12 cycles")
+check("⚠️ ...AND THE SITE DETECTOR IS NEVER ARMED. Not 'did not trip', and not "
+      "'its window stayed closed' - NEVER CALLED, counted on the object over "
+      "twelve cycles. That is what says the second detector is ABSENT at a "
+      "roster of one rather than merely quiet, and arming it would be two "
+      "latches and two banners for one fault with a false-positive record "
+      "measured on ONE comparison spent twice",
+      _n1_site_calls == []
+      and not _c1s.site_sign_wd.tripped
+      and _c1s.site_sign_wd.disagree_since is None)
+check("...the banner is the sentence that shipped",
+      _c1s.status_payload()["banner"]
+      == "FAILSAFE — the command opposed the meter. Inverter held at 0 W.")
+check("...the status page carries ONE sign row, with no unit tag and no site "
+      "row - an extra row for an identically-duplicated question is a page "
+      "advertising two guards where the install has one",
+      len([ck for ck in _c1s.checks()
+           if "sign watchdog" in ck["text"] or "SIGN-DISAGREEMENT" in ck["text"]
+           ]) == 1
+      and not any("site sign" in ck["text"].lower() for ck in _c1s.checks()))
+check("...and no `sign_site` claim can be filed at a roster of one, so the "
+      "arbitration is byte-identical to what shipped",
+      "sign_site" not in _c1s.reason)
+
+
+print()
+print("8. SAFETY-32 ROUND 4's ENUMERATION: the two `any(...)` reductions over")
+print("   the roster whose only evidence is a latched NON-FIRST unit")
+# ⚠️ THIS SECTION IS NOT SAFETY-32's SUBJECT, AND IT IS HERE BECAUSE OF IT.
+# SAFETY-32 was rejected three times for the same shape one step over: a
+# reduction over roster length whose fixture could not separate it from a slice
+# of the roster. Round 4 therefore ENUMERATED every such reduction in
+# `Controller` - min, max, sum, any, all, sorted, [0], [-1] over
+# `self.inverters` / `self.packs` - and mutated each to `[:1]` to find out which
+# were GRADED. Two survived a green 19-file suite, both `any(i.sign_wd.tripped
+# for i in self.inverters)`, both SAFETY-30's, and this block is what closes
+# them. No shipped behaviour changes; the reductions were already right.
+#
+#   main.py `cycle()`, the stale-input hold  -  +4 bytes, 19/19 GREEN
+#   main.py `publish()`, the status vocabulary -  +4 bytes, 19/19 GREEN
+#
+# ⚠️ WHY THE SUITE COULD NOT SEE THEM, and it is the same sentence every time:
+# the only fixture that reached either was a roster of ONE, where
+# `self.inverters[:1]` IS `self.inverters`. Part 3 above builds a two-inverter
+# roster in which ONLY inverter B is latched - the exact data both reductions
+# needed - and neither of them was read on it. A fixture of N units cannot
+# separate reductions that agree on N units, one card over.
+#
+# ⚠️ AND BOTH ORDERS ARE RUN, for 12g's reason: a single "only B is latched"
+# fixture kills `[:1]` and leaves `[-1:]` - grade the roster by its LAST unit -
+# answering identically. The pair is what leaves no single-position slice.
+
+
+async def _latched_then_input_gap(reversed_idx):
+    """Latch ONE unit's sign watchdog, then take a SIBLING's reading away.
+
+    ⚠️ A REAL LATCH FROM A REAL EPISODE, not `sign_wd.tripped = True`: the same
+    `_SignHass` plant and the same `_drive()` part 3 uses, so the detector
+    reaches its verdict through the shipped path. Then one pack's SoC goes
+    missing for LESS than `stale_input_s`, which is the state the hold exists
+    for - a single missed poll must not become a fault.
+
+    THE FAULT THIS GRADES: a LATCHED unit must take the whole roster past that
+    hold, because the unit's `failsafe` SET is the thing parking it. Return
+    early and the unit's last share stays standing for the heartbeat to
+    re-emit - main.py's own comment: "the trip would stop acting the moment a
+    poll was missed". A reduction that looked at one roster POSITION would do
+    exactly that whenever the latched unit was not the position it looked at.
+    """
+    hass = _SignHass(load_w=2000.0, reversed_units={reversed_idx})
+    c = _sign_site(hass)
+    await _drive(c)
+    sibling = c.inverters[1 - reversed_idx]
+    sibling.soc = None
+    c.inputs_bad_since = datetime.now(timezone.utc)
+    await c.cycle()
+    return c
+
+
+_gap_b = asyncio.run(_latched_then_input_gap(1))
+_gap_a = asyncio.run(_latched_then_input_gap(0))
+for _tag, _c, _i in (("B latched", _gap_b, 1), ("A latched", _gap_a, 0)):
+    _pl = [p for p in (_c.allocation.plans if _c.allocation else ())
+           if p.unit == _c.inverters[_i].name]
+    print(f"       {_tag}, sibling's SoC missing 0 s: inputs_ok="
+          f"{_c.inputs_ok}, reason={_c.reason!r}, latched unit target="
+          f"{_c.inverters[_i].target}, why={_pl[0].why if _pl else '(no plan)'}")
+check("PRECONDITION (habit 1): on BOTH runs the site really is in the state "
+      "the hold is for - one pack's reading is missing, `inputs_ok` is False, "
+      "the gap is younger than `stale_input_s`, and the SITE latch is NOT "
+      "tripped, so the only thing that can carry the roster past the hold is "
+      "the per-unit `any()`. And exactly ONE unit is latched on each run, at "
+      "the OTHER position",
+      _gap_b.inputs_ok is False and _gap_a.inputs_ok is False
+      and _gap_b.site_sign_wd.tripped is False
+      and _gap_a.site_sign_wd.tripped is False
+      and [i.sign_wd.tripped for i in _gap_b.inverters] == [False, True]
+      and [i.sign_wd.tripped for i in _gap_a.inverters] == [True, False])
+check("⚠️ A LATCHED UNIT TAKES THE ROSTER PAST THE STALE-INPUT HOLD FROM EITHER "
+      "POSITION. Neither cycle parks on `inputs missing`; both go on to file "
+      "claims, and the latched unit is commanded 0 W by its OWN `failsafe` "
+      "claim, named in the allocation record. Grade this with "
+      "`self.inverters[:1]` and the B-latched run returns early - the +4-byte "
+      "mutation that was GREEN across all 19 files - and with "
+      "`self.inverters[-1:]` the A-latched run does. The remedy stops acting "
+      "the moment a poll is missed, which is main.py's own sentence for it",
+      not (_gap_b.reason or "").startswith("inputs missing")
+      and not (_gap_a.reason or "").startswith("inputs missing")
+      and _gap_b.inverters[1].target == 0.0
+      and _gap_a.inverters[0].target == 0.0
+      and all(any(p.unit == c.inverters[i].name and "failsafe" in p.why
+                  for p in (c.allocation.plans if c.allocation else ()))
+              for c, i in ((_gap_b, 1), (_gap_a, 0))))
+# ...and the counterfactual, which is what makes the two lines above a
+# DIFFERENCE rather than a description: with NOTHING latched, the very same
+# input gap DOES park the site on the hold. Without this the checks above would
+# be satisfied by a `cycle()` that never had a hold at all.
+async def _gap_no_latch():
+    c = _sign_site(_SignHass(load_w=2000.0))
+    await _drive(c, n=2)
+    c.inverters[0].soc = None
+    c.inputs_bad_since = datetime.now(timezone.utc)
+    await c.cycle()
+    return c
+
+
+_gap_none = asyncio.run(_gap_no_latch())
+print(f"       nothing latched, same gap: reason={_gap_none.reason!r}")
+check("...and the counterfactual (habit 7): with NO unit latched the SAME gap "
+      "DOES hold the site - `inputs missing 0s` - so the two runs above show "
+      "the latch carrying the roster past a hold that is really there, rather "
+      "than a hold that never fires",
+      (_gap_none.reason or "").startswith("inputs missing")
+      and not any(i.sign_wd.tripped for i in _gap_none.inverters))
+
+# --- and the SECOND survivor: the status vocabulary publish() hands the broker
+# ⚠️ THE EXISTING CHECK IS AT A ROSTER OF ONE, which is why this one exists.
+# "MQTT's controller-status sensor reports 'failsafe'" above drives a
+# single-inverter Controller, where `self.inverters[:1]` is the whole roster.
+# A site at 0 W reporting `running` is the failure main.py's comment names, and
+# at two inverters it is reachable whenever the latched unit is not the one a
+# narrowed reduction happens to read.
+_pub = {}
+
+
+def _published(ctl):
+    _pub.clear()
+    ctl.mqtt = type("P", (), {"enabled": True,
+                              "publish": lambda _s, v: _pub.update(v),
+                              "close": lambda _s: None})()
+    ctl.publish()
+    return dict(_pub)
+
+
+_pb, _pa = _published(_gap_b), _published(_gap_a)
+_pn = _published(_gap_none)
+print(f"       published status: B latched -> {_pb.get('status')!r}, "
+      f"A latched -> {_pa.get('status')!r}, none latched -> "
+      f"{_pn.get('status')!r}")
+check("⚠️ AND `publish()` SAYS `failsafe` WHICHEVER UNIT IS LATCHED. The status "
+      "sensor is the vocabulary every dashboard already reads, and a site with "
+      "one inverter parked by its own failsafe claim reporting `running` is "
+      "the exact failure main.py's comment names. `self.inverters[:1]` "
+      "publishes `running` on the B-latched site - +4 bytes, GREEN across all "
+      "19 files - and `self.inverters[-1:]` does it on the A-latched one",
+      _pb.get("status") == "failsafe" and _pa.get("status") == "failsafe")
+check("...and the counterfactual: the same two-inverter site with NOTHING "
+      "latched publishes `running`, so the two above are a difference the "
+      "reduction produced and not a constant",
+      _pn.get("status") == "running" and _gap_none.auto is True)
+
+# --- and the THIRD one the same enumeration found: the OPERATOR'S BANNER -----
+# ⚠️ `status_payload()`'s failsafe branch is a fourth `any(... for i in
+# self.inverters)`, and `self.inverters[:1]` there was +4 bytes and GREEN across
+# all 19 files too - for the third time, because the only fixture that reached
+# it was a roster of ONE. The banner is the ONE sentence at the top of the page;
+# a site holding an inverter at 0 W under a "Stopped" or "NOT READY" headline
+# sends the operator to the wrong question.
+_bn_b = _gap_b.status_payload()
+_bn_a = _gap_a.status_payload()
+_bn_n = _gap_none.status_payload()
+print(f"       banner: B latched -> {_bn_b['banner']!r}")
+print(f"               A latched -> {_bn_a['banner']!r}")
+print(f"               none      -> {_bn_n['banner']!r}")
+check("⚠️ AND THE BANNER LEADS WITH THE FAILSAFE WHICHEVER UNIT IS LATCHED, at "
+      "level `bad`, with the N>=2 sentence that names the unit's OWN measured "
+      "power rather than the meter. `self.inverters[:1]` drops the headline on "
+      "the B-latched site and `self.inverters[-1:]` on the A-latched one, and "
+      "the page then leads with whatever the next branch says",
+      _bn_b["level"] == _bn_a["level"] == "bad"
+      and _bn_b["banner"] == _bn_a["banner"]
+      and _bn_b["banner"].startswith("FAILSAFE")
+      and "its own measured power" in _bn_b["banner"])
+check("...and the counterfactual again: the same site with nothing latched "
+      "carries neither that level nor that sentence, so the banner really is "
+      "being produced by the reduction under test",
+      not _bn_n["banner"].startswith("FAILSAFE")
+      and _bn_n["banner"] != _bn_b["banner"])
+
 print()
 if fails:
     print(f"{len(fails)} FAILED: {', '.join(fails)}")

goodwe_controller/tools/supervisor_shape_probe.py

@@ -40,8 +40,24 @@ DECISIONS/CFG-04-inverter-roster.md:
      reconstruction happens to lack FIRST, which is an artefact of the
      reconstruction and NOT evidence about `inverters`. B is that evidence.
   C. A two-inverter blob validates, per-entry types enforced.
-  D. An entry missing `name`, `setpoint_entity` or `max_w` is refused BY NAME,
-     and a JSON string where the list belongs is refused outright.
+  D. An entry missing `name`, `setpoint_entity`, `max_w` or `bms_soc_floor` is
+     refused BY NAME - one key varied from an ACCEPTED baseline, and the key's
+     own name asserted in the refusal - and a JSON string where the list belongs
+     is refused outright.
+
+⚠️ RE-RUN 2026-09-04 FOR SAFETY-32, which made `bms_soc_floor` a REQUIRED
+per-entry key. That change broke §C and, worse, left §D printing PASS on a
+refusal about the WRONG KEY - the exact decay MUTATION.md habit 8 describes. §C
+now carries the key and §D is an isolated counterfactual asserting the key name.
+
+⚠️ RE-RUN AGAIN 2026-09-05, on the same 2026.08.0 image, because SAFETY-32's
+rebase onto SPIKE-04a merged two independent edits into §C and §D - the required
+`bms_soc_floor` and inverter B's datasheet `max_w: 4600`. A transcript stitched
+together by hand from two runs is not a measurement, so this one was measured:
+`all cases passed`, exit 0. B's 4600 W is STILL its DATASHEET figure and not a
+reading off any unit, and B's `bms_soc_floor: 15` in §C is illustrative - it
+differs from A's 10 so that the per-entry number is exercised at all, while
+production is expected to have BOTH packs at 10 %.
 """
 import sys, json, copy
 sys.path.insert(0, "/usr/src/supervisor")
@@ -145,13 +161,47 @@ show("the same blob MERGED with these defaults - what App.options actually is",
      f"max_w={out.get('max_w')!r}" if ok else out)
 
 print("\n== C. AC2: a two-inverter blob an operator would POST ==")
+# ⚠️ B's 4600 W IS FROM ITS DATASHEET, NOT A PLACEHOLDER AND NOT A MEASUREMENT.
+# SPIKE-04a (2026-09-04) identified inverter B from the owner's model number and
+# serial - a GoodWe GW5048D-ES, serial `55048ESU22AW0992` - and the datasheet's
+# Max. Charging / Discharging Power is 4600 W. The 3000 W that stood here was a
+# hypothesis from before the unit was known and was simply wrong; the 5000 VA AC
+# nameplate is wrong for this purpose too, because nothing will be connected to
+# B's MPPTs so the battery path is the ceiling. See goodwe/NOTES.md §1.1b.
+#
+# ⚠️ AND IT IS STILL NOT MEASURED. The datasheet prints 4600 W under BATTERY
+# INPUT DATA, so it is probably DC, while `max_w` is graded against 0x0518 - the
+# inverter's own AC power. If it is DC the AC ceiling is nearer 4600 x 0.94 =
+# 4324 W. NOTES.md §1.1b records both readings without picking one, and
+# SAFETY-20's `min(max_w, dev_max)` is what bounds it correctly once dev_max is
+# read off the unit. So this blob is what an operator would POST TODAY, not a
+# claim about what inverter B will turn out to do.
+#
+# ⚠️ THE TWO RATINGS ARE STILL DIFFERENT (5000 / 4600) AND MUST STAY DIFFERENT.
+# This file's own §B is the record of what an instrument that cannot distinguish
+# its two cases is worth. Note that 5000/4600 is a much NARROWER separation than
+# the 5000/3000 it replaces - fine here, where nothing computes on the numbers,
+# and NOT fine in the test fixtures, which is why those keep a wide illustrative
+# pair instead of the real one.
+#
+# ⚠️ AND `bms_soc_floor` IS A REQUIRED PER-ENTRY KEY SINCE SAFETY-32, so this
+# blob is also what proves an operator can still POST a two-inverter config
+# after that. THE TWO FLOORS DIFFER ON PURPOSE (10 and 15): equal values would
+# not exercise a per-entry number at all, for the same reason the two ratings
+# must stay different.
+# ⚠️ NEITHER OF B'S NUMBERS IS A MEASUREMENT OF A PACK. Only inverter A's 10 %
+# is read off hardware (register 0x0560 on the reference GW5000S-BP). B's 4600 W
+# is its DATASHEET figure and B's 15 % is an illustrative value chosen to differ
+# - there is no second pack on this site, and in production BOTH packs are
+# expected to floor at 10 %, which is why the differing pair lives here and in
+# the fixtures rather than in config.yaml's example.
 TWO = dict(LIVE, setpoint_entity="", inverters=[
     {"name": "inverter-a", "setpoint_entity": "number.goodwe_a_setpoint_w",
      "soc_entity": "sensor.goodwe_a_soc", "batt_entity": "sensor.goodwe_a_battery_power",
-     "max_w": 5000},
+     "max_w": 5000, "bms_soc_floor": 10},
     {"name": "inverter-b", "setpoint_entity": "number.goodwe_b_setpoint_w",
      "soc_entity": "sensor.goodwe_b_soc", "batt_entity": "sensor.goodwe_b_battery_power",
-     "max_w": 3000}])
+     "max_w": 4600, "bms_soc_floor": 15}])
 ok, out = validate(MERGER.merge(copy.deepcopy(defaults), copy.deepcopy(TWO)))
 show("two inverters", ok, json.dumps(out.get("inverters")) if ok else out)
 # The validated blob, for anyone who wants to POST it. Optional: the probe's
@@ -162,14 +212,38 @@ except OSError:
     pass
 
 print("\n== D. what the schema itself refuses, per entry ==")
-for title, entry in (
-        ("no name", {"setpoint_entity": "number.b", "max_w": 3000}),
-        ("no setpoint_entity", {"name": "inverter-b", "max_w": 3000}),
-        ("no max_w", {"name": "inverter-b", "setpoint_entity": "number.b"}),
-        ("max_w over the range", {"name": "inverter-b", "setpoint_entity": "number.b", "max_w": 9999})):
-    ok, out = validate(dict(TWO, inverters=[entry]))
-    show(f"an entry with {title} is refused", not ok, out)
-ok, out = validate(dict(TWO, inverters="[]"))
+# ⚠️ ONE KEY VARIED FROM AN ACCEPTED BASELINE, AND THE REFUSAL IS ASSERTED BY
+# NAME - §B's two lessons, applied here because SAFETY-32 is what broke this
+# section. These entries used to omit several keys at once and assert a bare
+# `not ok`; the moment `bms_soc_floor` became required, the "no max_w" case went
+# on printing PASS while the refusal it printed said `Missing option
+# 'bms_soc_floor'`. A refusal of unknown provenance is not a measurement of
+# anything, so the baseline below is COMPLETE and each case deletes exactly one
+# key - or, for the range case, moves exactly one value out of range.
+D_FULL = {"name": "inverter-b", "setpoint_entity": "number.b", "max_w": 4600,
+          "bms_soc_floor": 10}
+# MERGED with the shipped defaults, exactly as §C is and as App.options is: the
+# raw LIVE blob is missing root keys unrelated to this section (`maintenance_
+# start_hour` on the run that caught it), and a baseline refused for one of THOSE
+# is the provenance problem §B2 is kept as a warning about.
+D_BASE = MERGER.merge(copy.deepcopy(defaults), copy.deepcopy(TWO))
+ok, out = validate(dict(D_BASE, inverters=[D_FULL]))
+show("the per-entry BASELINE is ACCEPTED, so every refusal below has nowhere "
+     "else to come from", ok, json.dumps(out.get("inverters")) if ok else out)
+for title, entry, expect in (
+        ("no name", {k: v for k, v in D_FULL.items() if k != "name"}, "name"),
+        ("no setpoint_entity",
+         {k: v for k, v in D_FULL.items() if k != "setpoint_entity"}, "setpoint_entity"),
+        ("no max_w", {k: v for k, v in D_FULL.items() if k != "max_w"}, "max_w"),
+        ("no bms_soc_floor (SAFETY-32)",
+         {k: v for k, v in D_FULL.items() if k != "bms_soc_floor"}, "bms_soc_floor"),
+        ("max_w over the range", dict(D_FULL, max_w=9999), "at most 5000"),
+        ("bms_soc_floor over the range (SAFETY-32)",
+         dict(D_FULL, bms_soc_floor=99), "at most 50")):
+    ok, out = validate(dict(D_BASE, inverters=[entry]))
+    show(f"an entry with {title} is refused, and the refusal NAMES it",
+         (not ok) and expect in str(out), out)
+ok, out = validate(dict(D_BASE, inverters="[]"))
 show("a JSON STRING where the list belongs is refused", not ok, out)