diff --git a/goodwe_controller/DECISIONS/SAFETY-136-slow-meter-settle.md b/goodwe_controller/DECISIONS/SAFETY-136-slow-meter-settle.md
index 27b1207..4818202 100644
--- a/goodwe_controller/DECISIONS/SAFETY-136-slow-meter-settle.md
+++ b/goodwe_controller/DECISIONS/SAFETY-136-slow-meter-settle.md
@@ -18,7 +18,7 @@ limit of SAFETY-128's fast meter fit). It counts arrivals, not changes of
 value: on the recorded day at 1-2 s the value-based test (`MeterFit.slow()`)
 called 24 % of the cycles slow, because the recorded house repeats a value
 for 5 s or more, and the commands then differed from base. So there is no
-option. At a 1-3 s meter it is false, the tuning is never touched and the
+option. At a meter with a usual gap of 3.75 s or less (1-3 s) it is false, the tuning is never touched and the
 commands are those of the tree before. A meter that Home Assistant passes on
 only when its value changes looks as slow as its changes are far apart, as it
 does to SAFETY-128's fits: a 1-2 s meter with a value that holds for 5 s or
@@ -63,10 +63,18 @@ from base. Together their export is 10 363 -> 9 559 Wh (-8 %) and import
 8 389 -> 7 835 Wh (-7 %). Only 5 export more than 1 Wh and 14 import more:
 export by at most 11.5 Wh (a 10 s battery lag, +10 %), import by at most
 17 Wh (a 12 s dead time, one inverter, +7 %), and the quarter average by at
-most 35 W, once 73 W (10 s meter, 10 s dead time, export -59 Wh). Slow
+most 35 W, except two import quarters: +73 W (10 s meter, 10 s dead time,
+export -59 Wh) and +40 W (`dead10_p6`, -6 to 34 W, export -68 Wh). Slow
 batteries with a dead time of 8-10 s export 5-39 % less; a 12 s dead time is
 the same. `READINGS_LATE` holds are the same seconds in every run.
 
+**Live, a 1-2 s meter can be judged slow.** Through Home Assistant on change,
+a value that holds for 4-10 s makes a 1 s meter slow. In 20 rig scenarios the
+commands were those of base. On the recorded day (1 s, N=1, 1 day, repeats
+dropped; its values hold about 5 s) it was slow in 33 % of cycles, half gain
+in 83 cycles, 15 422 commands in both trees, 4 016 with another value (at
+most 190 W apart), cost 0.61 EUR in both (review of SAFETY-136).
+
 **Does not change.** Meters of 1-2 s: the same commands in all 63 runs (34
 at 1 s, 29 at 2 s), and the recorded-day digests (see the report). The late
 verdict, the hold and the freeze tests. 15 s late is only partly helped, and
diff --git a/goodwe_controller/app/settle.py b/goodwe_controller/app/settle.py
index 04e202a..41a7ab4 100644
--- a/goodwe_controller/app/settle.py
+++ b/goodwe_controller/app/settle.py
@@ -1,10 +1,10 @@
 """SAFETY-136: on a slow meter, follow the meter gently until the last move shows.
 
-A meter whose readings come 4 s or more apart, read through Home Assistant, can
-be 10-15 s late. The loop then answers the same error again at each cycle
+A meter whose usual gap between readings is over 3.75 s (a 4-5 s meter), read
+through Home Assistant, can be 10-15 s late. The loop then answers the same error again at each cycle
 and overshoots. After a move of MOVE_W or more the loop therefore runs at
 WAIT_SHARE of its gain until the meter has fallen SHOWN of the move, or for at
-most WAIT_CYCLES cycles. Off for a meter under 4 s, so those sites run as
+most WAIT_CYCLES cycles. Off for a meter with a usual gap of 3.75 s or less, so those sites run as
 before. DECISIONS/SAFETY-136-slow-meter-settle.md.
 """
 
diff --git a/goodwe_controller/test_safety136.py b/goodwe_controller/test_safety136.py
index bca5545..5f3a957 100644
--- a/goodwe_controller/test_safety136.py
+++ b/goodwe_controller/test_safety136.py
@@ -118,6 +118,14 @@ check("the meter fell 200 W of 600 W: still half the gain",
 check("two cycles waited: the third has the whole gain",
       s.tuning(T, 300.0, True) is T and s.pend is None)
 s.moved(500.0, 0.0, 600.0, True)
+check("the meter went the other way by half the move: the wait stands",
+      s.tuning(T, 800.0, True).gain == T.gain * 0.5)
+s.reset()
+s.moved(500.0, 0.0, -600.0, True)
+check("a move down, the meter rose half the move: the whole gain",
+      s.tuning(T, 800.0, True) is T)
+s.reset()
+s.moved(500.0, 0.0, 600.0, True)
 check("the meter fell half the move: the whole gain at once",
       s.tuning(T, 200.0, True) is T and s.pend is None)
 s.moved(500.0, 0.0, -600.0, True)
@@ -148,6 +156,38 @@ print(f"    sum: {tot_on:.1f} Wh, off {tot_off:.1f} Wh")
 check("the four together export at most 65 % of the run with it off",
       tot_on <= 0.65 * tot_off)
 
+# --------------------------------------------------------------------------- #
+print("\n2b. Another writer overrides the loop: the wait ends")
+halved = []
+real_tuning = settle.Settle.tuning
+
+
+def watch_tuning(self, tuning, grid_w, slow):
+    r = real_tuning(self, tuning, grid_w, slow)
+    halved.append(r is not tuning)
+    return r
+
+
+settle.Settle.tuning = watch_tuning
+try:
+    mark = []
+
+    def bump(rig, c, t):
+        # right after a cycle that started a wait, another writer moves the
+        # command, so the next cycle sees `target != loop_w`
+        if not mark and c.settle.pend and t > 330.0:
+            mark.append(len(halved))
+            c.target = c.target + 300.0
+
+    r = R.Rig(house=oven, noise=80.0, p1_s=5.0, lag_s=10.0, lag_on=ON,
+              lag_off=900.0)
+    r.run(500.0, bump)
+finally:
+    settle.Settle.tuning = real_tuning
+nxt = nth(halved, mark[0] if mark else 10 ** 6)
+check("the cycle after an override has the whole gain",
+      bool(mark) and present(nxt) and nxt is False)
+
 # --------------------------------------------------------------------------- #
 print("\n3. 5 s meter, no lag")
 for name in ("oven", "kettle", "step", "stepup"):
@@ -202,15 +242,19 @@ for p1, expect in ((1.0, False), (2.0, False), (3.0, False), (4.0, True),
     last = nth(seen, -1)
     check(f"{p1:.0f} s meter: slow is {expect}",
           present(last) and last is expect)
-# A 1 s meter on a house that changes every 5 s repeats its value 4 times in 5:
-# the fit that counts changed values calls that slow, the readings do not (the
-# recorded day at 1-2 s showed it).
-s, m = settle.Settle(), MeterFit()
+# A 1 s meter on a house that changes every 5 s repeats its value 4 times in 5.
+# Where every reading arrives (the rig, a polled meter) it is not slow; where
+# Home Assistant sends a reading only when the value changes (ha_signed) the
+# same meter is slow (T3 of the review; DECISIONS).
+s, m, h = settle.Settle(), MeterFit(), settle.Settle()
 for k in range(60):
     s.arrive(float(k), k)
+    h.arrive(float(k), k // 5)
     m.add(float(k), 0.0, 1800.0 + 100.0 * (k // 5))
-check("a 1 s meter whose value changes every 5 s: not slow, though its value "
-      "repeats", not s.slow() and m.slow())
+check("every reading arrives, value changes every 5 s: not slow (the value "
+      "fit says slow)", not s.slow() and m.slow())
+check("the same meter through Home Assistant on change (a reading per "
+      "change): slow", h.slow())
 
 print()
 if fails: