SAFETY-32 fix round: grade the reduction on packs whose floors DIFFER

AC 6 was not met for the reduction main.py actually hands decide(). The
reviewer removed the per-pack grading twice and the 19-file suite stayed green
both times. No shipped behaviour changes here - app/ is untouched. What changes
is what observes it.

WHY BOTH SURVIVED, and it is this repository's recurring defect landing on this
card's own Decision 2: 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/last are four spellings of one answer. The only check
using a cut-out ABOVE the EMS floor - 12c's `_n1_deep` - ran on a roster of ONE,
where every reduction agrees. So no check ever evaluated binding_soc_floor() on
two packs whose floors differ. And the degeneracy is not a defence: this PR's own
argument ("unreachable is a claim about today's rule ordering, and
validate_startup_config() is also sim/inject.py's entry point") justifies writing
the reduction and equally obliges grading it.

  * test_peakguard.py part 12c2 - NEW. binding_soc_floor() on a roster whose
    per-pack floors ALL DIFFER: four SIMULATED cut-outs 30/25/40/35 %, each
    strictly above peak_soc_floor, so max (40), min (25), first (30), last (35)
    and the mean (32.5) are five different numbers and the extremes sit at
    INTERIOR roster positions. Two packs structurally cannot separate those -
    there max==last and min==first - and neither can three; four is the smallest
    roster that can. Precondition is its own check (habit 1); the counterfactual
    raises ONE pack - the last, neither extreme - past the old maximum and the
    answer follows (habit 7); peak_soc_floor is then lifted above every cut-out
    to grade the other side of the conjunction.
  * test_peakguard.py part 12g - REWRITTEN as behaviour (habit 4). The old
    assertion was `"self.packs" in <the unparsed call site>`, which `self.packs[:1]`
    still satisfies. It now drives REAL Controller.cycle()s on a two-pack roster,
    both packs at the same SoC, and reads PeakDecision.soc_floor - the argument,
    echoed back by the shipped decide(). The cut-outs are the only thing varied,
    and the SAME PAIR is carried in BOTH ORDERS.
  * ⚠️ TWO MORE SURVIVORS, found by this round's own battery and recorded rather
    than quietly repaired. (a) The first cut of 12g varied only pack B's cut-out,
    so `self.packs[-1:]` (+5 bytes) agreed with the roster on both runs and stayed
    green across all 19 files; the swapped pair closes it, and between the two
    directions no single-position slice passes all three runs. (b)
    binding_soc_floor()'s EMPTY-roster branch rewritten to `return None`
    (-43 bytes) was green too - a defensive branch nothing in the suite reached.
    Both directions of it are pinned now.
  * wired_over_limit() gained a prepare(ctl) hook, inside its retry loop so every
    attempt grades the same site.
  * In-process mutation rows 22, 22b and 23 record all four survivors.

Nine further source mutations went through the harness on the fixed tree
(binding_soc_floor -> floors[0] / floors[-1], pack_floor_pct conjoining with
min, pack_permissions grading against packs[0], `any` -> `all` on the unknown
floor, Controller.packs reversed, decide(soc=) taking inverters[0].soc, and the
two survivors above). Seven were red on first run; the two survivors are now red
as well. Three of the nine have a ZERO byte delta and still went red, which is
the direction stale bytecode cannot fabricate - and the harness purged anyway.

MANDATORY PROOF, all in the CI image over a tree rsynced to a plain path and
bind-mounted :ro, bytecode entry count printed before and after the purge:
each of the four mutations now takes test_peakguard.py RED, and the unmutated
tree is 19/19 green (twice, plus test_peakguard.py five times in isolation -
card CI-06's flake did not reappear). The zero-byte one was proved to have RUN rather than
assumed: dis.get_instructions LOAD_GLOBALs were [float, pack_floor_pct, any,
min] and binding_soc_floor() returned 25.0 on the 12c2 roster where the shipped
code returns 40.0.

No measurement of inverter B appears anywhere. It does not physically exist.
Every second-pack and fourth-pack figure added here is SIMULATED and labelled at
its definition; only inverter A's 10 % is read off hardware. No hardware was
touched.

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

goodwe_controller/CHANGELOG.md

@@ -58,6 +58,51 @@ thing that stops. Rule 8's own failure mode, reached through rule 8.
   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.
 
 ⚠️ **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

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

@@ -81,6 +81,116 @@ 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.
+
 ## 4. ⚠️ THE HONEST LIMITATION, and it is the reason this record exists
 
 **Under any configuration rule 8 accepts, `binding_soc_floor()` answers

goodwe_controller/test_peakguard.py

@@ -1063,9 +1063,15 @@ check("...and the status page says NOT COVERED rather than 'idle' or a green "
 # Controller rather than left to flake. It is bounded at three attempts: a fourth
 # would mean something is wrong that is not a boundary.
 def wired_over_limit(db, want_floor_w, opts=None, store=None, batt_w=0.0,
-                     standing_w=None, soc=55.0):
+                     standing_w=None, soc=55.0, prepare=None):
     """A real Controller driven one real cycle over a real over-limit quarter.
 
+    `prepare(ctl)` runs after the Controller is built and before the telegrams
+    are fed, which is the only window in which a SECOND inverter can be put on
+    the roster and given its own readings. It is inside the retry loop on
+    purpose: each attempt builds a fresh Controller, so a hook that mutated only
+    the first one would leave attempts 2-4 grading a different site.
+
     ⚠️ THE LOAD IS DERIVED FROM WHERE IN THE QUARTER THE WALL CLOCK LANDS, and
     it has to be. `excess = 900 × (projected − ceiling) ÷ remaining`, so a fixed
     5200 W fixture asks for 1700 W with ten minutes left and 300 kW with five
@@ -1079,6 +1085,8 @@ def wired_over_limit(db, want_floor_w, opts=None, store=None, batt_w=0.0,
                         store=_Store(store) if store is not None else None)
         ctl.auto, ctl.soc, ctl.batt = True, soc, batt_w
         ctl.dev_min, ctl.dev_max = -LIVE_MAX_W, LIVE_MAX_W
+        if prepare is not None:
+            prepare(ctl)
         if standing_w is not None:
             # The command STANDING ON THE WIRE, which is what B3 below needs to
             # differ from the measured battery power.
@@ -3291,6 +3299,130 @@ check("...and on the pair rule 8 REFUSES - a cut-out above the EMS floor - the "
       "would be satisfied by `soc_floor` typed straight through",
       _n1_deep == B_PACK_FLOOR_DEEP_SIM and _n1_deep != SOC_FLOOR)
 
+# --- 12c2. THE REDUCTION ITSELF, ON A ROSTER WHOSE FLOORS ALL DIFFER ---------
+# ⚠️ THIS IS THE CHECK PR #48 DID NOT HAVE, AND ITS ABSENCE IS THIS REPOSITORY'S
+# RECURRING DEFECT LANDING ON THIS CARD. Every other fixture above grades
+# `binding_soc_floor()` on a roster of ONE (12c) or on a config rule 8 accepts,
+# where `pack_floor_pct()` answers `peak_soc_floor` for EVERY pack - so `floors`
+# is a list of IDENTICAL numbers and `max`, `min`, `floors[0]` and `floors[-1]`
+# are four spellings of one answer. A reviewer changed `max(floors)` to
+# `min(floors)` and all 19 files stayed GREEN. The data could not reach the
+# failure the check names.
+#
+# ⚠️ AND THE DEGENERACY IS NOT A DEFENCE, BY THIS CARD'S OWN ARGUMENT.
+# `pack_floor_pct()`'s docstring, main.py's call-site comment and DECISIONS §4
+# all say "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". That argument justifies WRITING the reduction; it equally OBLIGES
+# grading it. Via `sim/inject.py`, which does not run rule 8 at all, `min` here
+# lets the guard command a forced discharge down to 20 % on a pack whose BMS
+# cuts out at 35 %. So the reduction is graded on a roster rule 8 would refuse -
+# deliberately, because that is the only roster on which it does anything.
+#
+# ⚠️ THE FOUR NUMBERS, AND WHY THESE FOUR. Every one is ABOVE `peak_soc_floor`,
+# so `pack_floor_pct()` selects the CUT-OUT for each and `floors` IS this tuple
+# in this order. Read as a list, the four candidate reductions are four
+# DIFFERENT numbers:
+#
+#         floors      = (30, 25, 40, 35)
+#         max(floors) = 40   <- the shipped answer, the deepest cut-out
+#         min(floors) = 25
+#         floors[0]   = 30   <- "grade the site against pack A", survivor 2's
+#                              defect expressed inside the function
+#         floors[-1]  = 35
+#         mean        = 32.5
+#
+# so no two of them can be confused, and the extremes sit at INTERIOR roster
+# positions (index 2 is the max, index 1 the min) - which is what a two-pack
+# fixture structurally cannot do, since there `max` and `last` coincide and so
+# do `min` and `first`. Three packs cannot do it either. Four is the smallest
+# roster that separates all four.
+#
+# ⚠️ ALL FOUR ARE SIMULATED. There is ONE pack on this site; inverter B is ~8
+# days out and does not physically exist, and a roster of four certainly does
+# not. Nothing here is a measurement of anything. `binding_soc_floor()` is a
+# pure function over a list and these are the inputs that make its answer
+# observable - the only measured cut-out in this file is inverter A's 10 %.
+_MANY_FLOORS_SIM = (30.0, 25.0, 40.0, 35.0)
+_many_packs = [Pack(f"sim-{i}", 50.0, f) for i, f in enumerate(_MANY_FLOORS_SIM)]
+_many_derived = [pack_floor_pct(pk, SOC_FLOOR) for pk in _many_packs]
+_many_answers = {"max": max(_MANY_FLOORS_SIM), "min": min(_MANY_FLOORS_SIM),
+                 "first": _MANY_FLOORS_SIM[0], "last": _MANY_FLOORS_SIM[-1],
+                 "mean": sum(_MANY_FLOORS_SIM) / len(_MANY_FLOORS_SIM)}
+print(f"    a SIMULATED roster of {len(_many_packs)} with cut-outs "
+      f"{[f'{f:.0f}' for f in _MANY_FLOORS_SIM]} % over peak_soc_floor "
+      f"{SOC_FLOOR:.0f} % -> per-pack floors {[f'{f:.0f}' for f in _many_derived]} %")
+print("    candidate reductions: "
+      + ", ".join(f"{k}={v:g}" for k, v in _many_answers.items())
+      + f"; binding_soc_floor() -> {binding_soc_floor(_many_packs, SOC_FLOOR):g}")
+check("PRECONDITION (MUTATION.md habit 1): on this roster max, min, first, last "
+      "and the mean are FIVE DIFFERENT numbers, and every cut-out is strictly "
+      "above peak_soc_floor so the per-pack floors ARE the cut-outs. A later "
+      "edit that makes any two of them equal takes this line red rather than "
+      "leaving the claim below quietly vacuous - which is exactly what a "
+      "rule-8-valid roster does to it",
+      len(set(_many_answers.values())) == 5
+      and all(f > SOC_FLOOR for f in _MANY_FLOORS_SIM)
+      and _many_derived == list(_MANY_FLOORS_SIM))
+check("⚠️ `binding_soc_floor()` answers the MAXIMUM of the packs' own floors - "
+      "the DEEPEST cut-out on the roster - and on this roster that is not the "
+      "minimum, not the first pack's, not the last pack's and not the mean. "
+      "This is the conservative pairing with the MINIMUM in `pack_floor_soc`: "
+      "`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. `min(floors)` here would let a forced discharge be "
+      "sized down to 25 % on the pack whose BMS cuts out at 40 %",
+      binding_soc_floor(_many_packs, SOC_FLOOR) == _many_answers["max"]
+      and binding_soc_floor(_many_packs, SOC_FLOOR) != _many_answers["min"]
+      and binding_soc_floor(_many_packs, SOC_FLOOR) != _many_answers["first"]
+      and binding_soc_floor(_many_packs, SOC_FLOOR) != _many_answers["last"]
+      and binding_soc_floor(_many_packs, SOC_FLOOR) != _many_answers["mean"])
+# ⚠️ THE COUNTERFACTUAL, ONE THING VARIED FROM AN ACCEPTED BASELINE (habit 7).
+# The pack that is raised is the LAST one, which is neither the max nor the min
+# above - so this line can only move if the reduction really looked at it. It is
+# raised past the old maximum, and the answer follows it.
+_deeper = list(_many_packs)
+_deeper[-1] = Pack("sim-3", 50.0, 45.0)
+print(f"    ...raise ONLY the last pack 35 % -> 45 %: "
+      f"{binding_soc_floor(_many_packs, SOC_FLOOR):g} -> "
+      f"{binding_soc_floor(_deeper, SOC_FLOOR):g}")
+check("...and the counterfactual: raise ONE pack - the LAST, which was neither "
+      "the deepest nor the shallowest - past the old deepest, and the answer "
+      "follows it to 45 %. So the reduction reads every entry, and the 40 % "
+      "above was not a coincidence of position",
+      binding_soc_floor(_deeper, SOC_FLOOR) == 45.0
+      and binding_soc_floor(_many_packs, SOC_FLOOR) == 40.0)
+# ...and the EMS floor is still the other half of the conjunction: lift it above
+# every cut-out on the roster and it wins for all of them.
+_over = SOC_FLOOR + 40.0
+check("...and with `peak_soc_floor` lifted above every cut-out on this roster "
+      "the answer is the OPTION, for all four - `pack_floor_pct()` is a "
+      "conjunction and this is its other side, so a reduction that dropped the "
+      "EMS floor entirely goes red here as well as in the N=1 sweep",
+      _over > max(_MANY_FLOORS_SIM)
+      and binding_soc_floor(_many_packs, _over) == float(_over))
+
+# ...and the EMPTY roster, the SECOND thing this fix round's own battery found
+# unguarded: `return None` in place of `float(soc_floor)` there (-43 bytes) was
+# GREEN across all 19 files. It is a defensive branch - `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. Pinned
+# rather than argued, because a branch with no check is a branch that can be
+# rewritten under a green suite, and BOTH directions of it are asserted: the
+# option when there is one, and None when there is not.
+print(f"    an EMPTY roster: binding_soc_floor([], {SOC_FLOOR:.0f}) -> "
+      f"{binding_soc_floor([], SOC_FLOOR)}, binding_soc_floor([], None) -> "
+      f"{binding_soc_floor([], None)}")
+check("an EMPTY roster answers the OPTION - there is no pack to conjoin with, "
+      "so there is nothing to raise the floor above it - and answers None when "
+      "the option itself is None. ⚠️ AND IT CANNOT PERMIT ANYTHING EITHER WAY: "
+      "`Controller.pack_floor_soc` is None at an empty roster, and "
+      "`discharge_permitted()` refuses on a None SoC whatever the floor is",
+      binding_soc_floor([], SOC_FLOOR) == float(SOC_FLOOR)
+      and binding_soc_floor([], None) is None
+      and discharge_permitted(None, binding_soc_floor([], SOC_FLOOR))
+      == FLOOR_SOC_UNREADABLE)
+
 # --- 12d. AC 4: pack_floor_soc's None-on-missing is untouched ---------------
 _c32, _h32 = built(TMP / "s32.db")
 _b32, _restore32 = second_inverter(_c32)
@@ -3436,17 +3568,147 @@ check("...and with the key absent entirely the one pack takes the SHIPPED "
       and BMS_FLOOR == PEAK_OPTION_DEFAULTS["bms_soc_floor"])
 
 # --- 12g. the call site: main.py hands decide() the per-pack reduction -------
+# ⚠️ REWRITTEN. THIS CHECK USED TO BE A SUBSTRING TEST OVER main.py's SOURCE -
+# `"self.packs" in _floor_call` - AND IT WAS NOT A PIN. The mutation that grades
+# the whole site against pack A's floor is `self.packs[:1]`, which still
+# CONTAINS that substring: +4 bytes, all 19 files green, including this one.
+# MUTATION.md habit 4, in the same costume CFG-04's `if entity:` wore: the check
+# asserted against the TEXT of the thing under test instead of its behaviour.
+# The AST is still unparsed and PRINTED (habit 3) because a reader wants to see
+# the expression, but nothing below asserts on its spelling. Which OPTION KEY
+# the floor comes from is a claim genuinely about the code's shape and is pinned
+# on the source in part 4, where it already lives; what is graded here is the
+# VALUE that reaches `decide()`.
 _floor_call = ast.unparse(next(k.value for k in _call.keywords
                                if k.arg == "soc_floor"))
 print(f"    main.py's decide(soc_floor=...) is `{_floor_call}`")
-check("main.py hands `decide()` the PER-PACK reduction of `peak_soc_floor`, "
-      "not the bare option: `binding_soc_floor(self.packs, ...)`, so the floor "
-      "the site is graded against can never sit below the deepest cut-out on "
-      "the roster. The key and the one defaults mirror are still in that "
-      "expression - graded three parts above - because this card moved the "
-      "REDUCTION and not the source",
-      "binding_soc_floor" in _floor_call and "self.packs" in _floor_call
-      and '"peak_soc_floor"' in _floor_call.replace("'", '"'))
+
+
+def _floor_reaching_decide(tag, a_floor, b_floor, peak_floor=None):
+    """Drive ONE REAL `Controller.cycle()` on a TWO-pack roster and report the
+    floor `peakguard.decide()` was actually handed, read back off the decision
+    main.py itself filed (`PeakDecision.soc_floor` is the argument, echoed).
+
+    ⚠️ NOT A SPY AND NOT A GREP. The shipped `cycle()` runs, the shipped
+    `binding_soc_floor()` runs, and the observable is a field of the decision
+    the controller stored - so the whole path from `Controller.inverters` to the
+    guard's argument is under test. Both packs are held at the SAME SoC and the
+    same everything else; the only things that vary between calls are the packs'
+    own cut-outs.
+    """
+    opts = {} if peak_floor is None else {"peak_soc_floor": peak_floor}
+
+    def _prep(ctl):
+        second_inverter(ctl)
+        for inv, fl in zip(ctl.inverters, (a_floor, b_floor)):
+            inv.soc, inv.batt, inv.bms_soc_floor = SOC_BETWEEN, 0.0, fl
+
+    ctl, hs, _net, _rem = wired_over_limit(
+        TMP / f"s32g_{tag}.db", LIVE_MAX_W * 0.3, opts=opts, prepare=_prep)
+    return ctl, ctl.peak_guard
+
+
+# BASELINE: neither cut-out is above the EMS floor, so the conjunction selects
+# the option for both packs and the site floor is `peak_soc_floor`. At
+# SOC_BETWEEN (5 points above it) the guard is permitted and sizes a real
+# discharge. Both runs below vary one thing from THIS baseline (habit 7).
+_g_ctl_shallow, _g_shallow = _floor_reaching_decide(
+    "shallow", A_PACK_FLOOR, B_PACK_FLOOR_SIM)
+# The SECOND pack's cut-out moves above the EMS floor. Nothing else changes -
+# not pack A, not either SoC, not the option.
+_g_ctl_deep, _g_deep = _floor_reaching_decide(
+    "deepb", A_PACK_FLOOR, B_PACK_FLOOR_DEEP_SIM)
+# ⚠️ AND THE SAME TWO CUT-OUTS SWAPPED BETWEEN THE PACKS, which is the mirror
+# 12h already runs on the SoC axis and which THIS part needed and did not have.
+# Without it, `self.packs[-1:]` - grade the site against the LAST pack - answers
+# the same two numbers as the full roster on the two runs above, because pack A
+# is never the deeper one there. It SURVIVED the first cut of this fix round,
+# +5 bytes, all 19 files green. With the pair swapped, `[:1]` and `[-1:]`
+# disagree with the roster in opposite runs, and no single-position slice is
+# left that can pass all three.
+_g_ctl_deepa, _g_deepa = _floor_reaching_decide(
+    "deepa", B_PACK_FLOOR_DEEP_SIM, B_PACK_FLOOR_SIM)
+print(f"    REAL two-pack cycles, both packs at {SOC_BETWEEN:.0f} %, "
+      f"peak_soc_floor {SOC_FLOOR:.0f} % (every cut-out but inverter-a's own "
+      f"{A_PACK_FLOOR:.0f} % is SIMULATED):")
+print(f"      cut-outs A {A_PACK_FLOOR:.0f} / B {B_PACK_FLOOR_SIM:.0f} % -> "
+      f"decide() got soc_floor={_g_shallow.soc_floor:g} %, floor_limited="
+      f"{_g_shallow.floor_limited}")
+print(f"      cut-outs A {A_PACK_FLOOR:.0f} / B {B_PACK_FLOOR_DEEP_SIM:.0f} % -> "
+      f"decide() got soc_floor={_g_deep.soc_floor:g} %, floor_limited="
+      f"{_g_deep.floor_limited}")
+print(f"      ...the SAME PAIR SWAPPED, A {B_PACK_FLOOR_DEEP_SIM:.0f} / B "
+      f"{B_PACK_FLOOR_SIM:.0f} % -> decide() got "
+      f"soc_floor={_g_deepa.soc_floor:g} %, floor_limited="
+      f"{_g_deepa.floor_limited}")
+check("⚠️ THE CALL-SITE PIN, AS BEHAVIOUR. Change the cut-out of a NON-FIRST "
+      "pack and the floor that reaches `decide()` MOVES with it: 20 % (the "
+      "option, because B's cut-out is under it) becomes 35 % (B's own cut-out, "
+      "because it is over it). A call site that graded the site against "
+      "`self.packs[:1]` - pack A alone - answers 20 % both times and cannot "
+      "produce this difference, and neither can `min(floors)`; that is the "
+      "+4-byte mutation that walked through the substring test this check "
+      "replaces",
+      _g_shallow.soc_floor == float(SOC_FLOOR)
+      and _g_deep.soc_floor == float(B_PACK_FLOOR_DEEP_SIM)
+      and _g_deep.soc_floor != _g_shallow.soc_floor)
+check("⚠️ ...AND THE SAME PAIR SWAPPED GIVES THE SAME SITE FLOOR, so what is "
+      "reduced is the ROSTER and not a POSITION in it. `self.packs[-1:]` - the "
+      "LAST pack alone - agrees with the roster on both runs above and answers "
+      "20 % here, where the deeper cut-out sits on the FIRST pack. That "
+      "mutation SURVIVED the first cut of this fix round; this is the line "
+      "that closes it, and between them the two directions leave no "
+      "single-position slice that can pass all three runs",
+      _g_deepa.soc_floor == float(B_PACK_FLOOR_DEEP_SIM)
+      and _g_deepa.soc_floor == _g_deep.soc_floor
+      and _g_deepa.soc_floor != _g_shallow.soc_floor
+      and [pk.bms_floor for pk in _g_ctl_deepa.packs]
+      == [B_PACK_FLOOR_DEEP_SIM, B_PACK_FLOOR_SIM])
+check("...and it is not just a number in a report - THE SITE'S VERDICT FLIPS. "
+      "With both packs at 25 % the shallow site is permitted and the guard "
+      "sizes a positive forced discharge; BOTH deep sites are FLOOR-LIMITED, "
+      "because 25 % is at or below the 35 % cut-out wherever on the roster it "
+      "sits. Grading the roster against one position would spend a pack its "
+      "BMS stops - through `sim/inject.py`, which does not run rule 8 - and "
+      "this line is what goes red when it does",
+      _g_shallow.floor_limited is False and _g_shallow.floor_w > 0.0
+      and _g_deep.floor_limited is True and _g_deep.floor_w == 0.0
+      and _g_deep.soc_floor_reason == FLOOR_AT_SOC_FLOOR
+      and _g_deepa.floor_limited is True and _g_deepa.floor_w == 0.0
+      and _g_deepa.soc_floor_reason == FLOOR_AT_SOC_FLOOR)
+check("...and the precondition that makes the three runs comparable (habit 1): "
+      "the two deep runs carry THE SAME PAIR of cut-outs in opposite orders, "
+      "those two numbers really are different, and all six packs are at the "
+      "SAME state of charge - so neither the SoC reduction nor a roster "
+      "position can be what separated the answers",
+      B_PACK_FLOOR_SIM != B_PACK_FLOOR_DEEP_SIM
+      and A_PACK_FLOOR != B_PACK_FLOOR_DEEP_SIM
+      and sorted(pk.bms_floor for pk in _g_ctl_deepa.packs)
+      == sorted([B_PACK_FLOOR_SIM, B_PACK_FLOOR_DEEP_SIM])
+      and [pk.soc for pk in _g_ctl_deep.packs] == [SOC_BETWEEN, SOC_BETWEEN]
+      and [pk.soc for pk in _g_ctl_deepa.packs] == [SOC_BETWEEN, SOC_BETWEEN]
+      and [pk.bms_floor for pk in _g_ctl_deep.packs]
+      == [A_PACK_FLOOR, B_PACK_FLOOR_DEEP_SIM]
+      and [pk.bms_floor for pk in _g_ctl_shallow.packs]
+      == [A_PACK_FLOOR, B_PACK_FLOOR_SIM]
+      and _g_ctl_deep.pack_floor_soc == _g_ctl_shallow.pack_floor_soc
+      == _g_ctl_deepa.pack_floor_soc)
+# ...and the OTHER half of the conjunction still reaches the same argument, so
+# the reduction did not replace the source: move `peak_soc_floor` itself, with
+# both cut-outs held under it, and the floor follows the OPTION.
+_G_OPT_ODD = SOC_FLOOR + 22.0
+_, _g_opt = _floor_reaching_decide("opt", A_PACK_FLOOR, B_PACK_FLOOR_SIM,
+                                   peak_floor=_G_OPT_ODD)
+print(f"      peak_soc_floor moved to {_G_OPT_ODD:.0f} % with both cut-outs "
+      f"under it -> decide() got soc_floor={_g_opt.soc_floor:g} %")
+check("...and with both cut-outs held below it, moving `peak_soc_floor` moves "
+      "the floor that reaches `decide()` too - so this card moved the "
+      "REDUCTION and not the SOURCE. WHICH key that is, is a claim about the "
+      "code's shape and stays pinned on the source in part 4; that the option "
+      "still governs is a claim about behaviour and is graded here",
+      _g_opt.soc_floor == float(_G_OPT_ODD)
+      and _G_OPT_ODD > B_PACK_FLOOR_SIM and _G_OPT_ODD > A_PACK_FLOOR
+      and _g_opt.soc_floor != _g_shallow.soc_floor)
 
 # --- 12h. the operator-facing half, driven through the shipped method -------
 # ⚠️ RUN, NOT GREPPED. CFG-04's review found a substring test over main.py that
@@ -3729,6 +3991,61 @@ MUT.append(("the per-pack notification naming a fixed roster position",
             and "inverter-b" not in _body32b.split("WHICH PACK")[-1],
             "part 12h: the same two readings swapped between the packs"))
 
+# ---- the two PR #48 SURVIVORS, and why they survived -------------------------
+# ⚠️ BOTH OF THESE WERE APPLIED TO app/ AND ALL 19 FILES STAYED GREEN. They are
+# this repository's recurring defect - a check whose DATA cannot reach the
+# failure it names - landing on this card's own reduction, and they are recorded
+# here as rows so that a later edit which re-degenerates the fixtures goes red
+# on the row rather than on nothing.
+
+# 22. `binding_soc_floor()` reduced with `min()` instead of `max()`. ZERO byte
+#     delta, and the reviewer proved by disassembly that it really ran. It
+#     survived because 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. 12c2's four simulated cut-outs are the data that
+#     reaches it.
+_mut22 = [min(_MANY_FLOORS_SIM), _MANY_FLOORS_SIM[0], _MANY_FLOORS_SIM[-1],
+          sum(_MANY_FLOORS_SIM) / len(_MANY_FLOORS_SIM)]
+MUT.append(("`binding_soc_floor()` reduced with `min()`, `first` or `last` "
+            "instead of `max()` - PR #48 SURVIVOR 1, zero byte delta",
+            binding_soc_floor(_many_packs, SOC_FLOOR) == max(_MANY_FLOORS_SIM)
+            and all(binding_soc_floor(_many_packs, SOC_FLOOR) != m
+                    for m in _mut22),
+            "part 12c2: 'max, min, first, last and the mean are FIVE DIFFERENT "
+            "numbers'. ⚠️ The four cut-outs sit on a roster rule 8 would REFUSE, "
+            "deliberately: on one it accepts all four reductions are equal, "
+            "which is precisely why this mutation survived"))
+
+# 22b. `binding_soc_floor()`'s EMPTY-roster branch rewritten to `return None`.
+#      -43 bytes, green across all 19 files until 12c2's last block pinned it.
+#      A defensive branch nothing reaches is still a branch.
+MUT.append(("`binding_soc_floor([])` answering None instead of the option",
+            binding_soc_floor([], SOC_FLOOR) == float(SOC_FLOOR)
+            and binding_soc_floor([], None) is None,
+            "part 12c2's last block: 'an EMPTY roster answers the OPTION ... "
+            "and answers None when the option itself is None'"))
+
+# 23. the CALL SITE grading the whole site against ONE ROSTER POSITION -
+#     `self.packs[:1]` (+4 bytes, PR #48's survivor) or `self.packs[-1:]`
+#     (+5 bytes, which survived the FIRST CUT OF THIS FIX ROUND until the
+#     swapped pair was added). 12g's old substring assertion
+#     (`"self.packs" in _floor_call`) passes on both, because a slice still
+#     contains that substring. Three runs close both directions at once: the
+#     deep cut-out is on the SECOND pack in one and on the FIRST in another, and
+#     the site floor is the same 35 % in both.
+MUT.append(("main.py grading the site against ONE roster position - "
+            "`self.packs[:1]` (PR #48 SURVIVOR 2) or `self.packs[-1:]`",
+            _g_deep.soc_floor == float(B_PACK_FLOOR_DEEP_SIM)
+            and _g_deepa.soc_floor == float(B_PACK_FLOOR_DEEP_SIM)
+            and _g_shallow.soc_floor == float(SOC_FLOOR)
+            and _g_deep.floor_limited and _g_deepa.floor_limited
+            and not _g_shallow.floor_limited,
+            "part 12g, REWRITTEN as behaviour: real two-pack cycles, the SAME "
+            "pair of cut-outs carried in both orders, and both the floor "
+            "reaching decide() and the site's verdict follow the ROSTER rather "
+            "than a position. ⚠️ A substring test over main.py cannot see a "
+            "slice; MUTATION.md habit 4"))
+
 for label, caught, where in MUT:
     print(f"    {'caught' if caught else 'NOT CAUGHT':>10}  {label}")
     print(f"                {where}")
@@ -3748,7 +4065,18 @@ check("...and SAFETY-32's own eight defeats are each caught too - the flat "
       "first pack's floor borrowed for all, the reduction dropped, an absent "
       "cut-out defaulted to 0 %, rule 8 grading one unit, rule 8 naming a "
       "hardwired one, and the alarm naming a position instead of a pack",
-      all(m[1] for m in MUT[13:21]) and len(MUT) == 21)
+      all(m[1] for m in MUT[13:21]))
+check("...⚠️ AND SO ARE THE FOUR DEFEATS THAT SURVIVED A GREEN 19-FILE SUITE, "
+      "which is why this fix round exists: `min(floors)` in place of "
+      "`max(floors)` (zero byte delta) and `self.packs[:1]` at the call site "
+      "(+4 bytes), both green on d9388c8; `self.packs[-1:]` (+5 bytes), green "
+      "on the FIRST CUT OF THE FIX ITSELF; and the empty-roster branch "
+      "rewritten (-43 bytes). None of the four is a new guard - no shipped "
+      "behaviour changed - they are checks whose DATA could not reach the "
+      "failure they named, which is the defect this repository has now shipped "
+      "in six consecutive cards. The last two were found by this round's own "
+      "battery and are recorded rather than quietly repaired",
+      all(m[1] for m in MUT[21:24]) and len(MUT) == 24)
 check("...and so is the last one, which is not a defeat at all but the EDGE OF "
       "THIS CARD: the grid-following loop can still take the pack past this "
       "floor, because the floor bounds the peak guard's own forced discharge "