memory: snapshot — U4 released at v0.4.0, next unit undecided

Current state rewritten for the post-U4 position: 410 tests, v0.4.0 tagged,
tree not pushed, all three gates closed. Carries the session's U3 recommendation
with its three grounds AND its counter-argument, so the operator can take the
call without reloading the unit.

The methodology-proposals row goes from three to four and is now marked
explicitly untracked by operator choice — the new one is the contract-time
vacuity pass, which is the only one of the four with measured evidence behind
it after five of seven U4 falsifiers turned out not to discriminate.

persistent-memory.md

@@ -23,8 +23,9 @@ _As of 2026-09-22:_
   **U1 → U2 → {U3, U4, U5} → U7**, with **U6 independent**.
 - **U1, U2, U4 and U5 are landed.** U1 `ce598b3`; U2 `c7f9437` → `v0.2.0`,
   `5e41108` → `v0.2.1`, `026a1fc` → `v0.2.2`; U5 `c015a91` + `95beede` →
-  `v0.3.0`. **U4 landed 2026-09-22** — 396 tests green (341 → 396), deployed and
+  `v0.3.0`. **U4 landed 2026-09-22** — 410 tests green (341 → 410), deployed and
-  verified live, 24/24 booth pages 200, layout probe clean.
+  verified live, 25/25 booth pages 200, layout probe clean. Tree clean at
+  `8f81d8f`; NOT PUSHED (push is the operator's call and he has not given it).
 - **U4 released as `v0.4.0`** (operator approved the minor on 2026-09-22).
   `c3a97c1` is the unit; the release commit carries the pre-existing fixes the
   bug-hunt panel surfaced in touched files. The tag waited for the last gate to
@@ -34,12 +35,20 @@ _As of 2026-09-22:_
   deliberate and it CHANGES HOW THE 2026-10-06 RE-COUNT READS: the hold rides
   for free, but not-pressing-`keep` has to be learned, so a flat `.forever` rate
   does not falsify anything. Read its entry before measuring.
-- **THE NEXT UNIT IS THE OPERATOR'S CALL.** U3 (declared embed seam) and U6
+- **THE NEXT UNIT IS THE OPERATOR'S CALL and has not been made.** U3 (declared
-  (benches) are both unblocked; U7 waits on the rest. U6 is independent of
+  embed seam) and U6 (benches) are both unblocked; U7 waits on the rest.
-  everything and was conceptually unblocked by U5 giving job 5 a home; U3 is
+  **The session's recommendation is U3**, on three grounds that do not need
-  where verbatim-booth provenance was deferred to, and U4 added a fourth reason
+  re-deriving: (1) U7 depends on {U3, U4, U5}, so U3 is the only remaining unit
-  to want it — a verbatim booth has no Booth-rendered header, so its lifetime
+  on the critical path to the endgame — U6 is independent and can land any time;
-  line lives only on the index card and the marks page.
+  (2) U4 added a fourth reason to want it, because a verbatim booth has no
+  Booth-rendered header and so carries its lifetime line only on the index card
+  and the marks page; (3) it is the biggest DELETER in the plan — `inline.py`
+  entire, six regexes against arbitrary author HTML, two floating chips and a
+  route — and regex-injecting into author HTML is named in the IA doc as the
+  single most fragile thing in the service. ⚠ The counter-argument, stated
+  because it is real: U3 is also the RISKIEST unit, since it changes how every
+  verbatim booth renders and verbatim booths are the operator's own reports.
+  U6 is the low-risk session and closes the biggest single number (69% rot).
 - **No gate is outstanding.** All three ran on U4 and were folded in: the
   `/heid-contract-review` panel (`01M34VX0SH23Y3VC92E7GM4S70`), the
   `/heid-code-review` panel (`01M34WAFJC3RTERFYBBZJN1SVG`) and the
@@ -53,14 +62,20 @@ _As of 2026-09-22:_
   U4's success criterion. ⚠ Only 4 booths carry marks at all, so the hold's live
   blast radius is small and the prediction rests on both halves of U4 — see its
   entry for what a null result would and would not mean.
-- **Three methodology proposals from this session sit with the operator**, routed
+- **FOUR methodology proposals sit with the operator, all UNTRACKED BY OPERATOR
-  by heid rather than decided unilaterally: reshaping the paraphrase gate toward
+  CHOICE** (no issue, no ticket — they are `/heid*` skill changes, not this
-  a drift-check for narrative-heavy contracts, a standing
+  repo's work, and are recorded here only so they are not lost). Three are from
-  "green-tests-prove-nothing" direction for the code-review gate, and regin's
+  the U5 round: reshaping the paraphrase gate toward a drift-check for
-  table-vs-signature consistency pass. They are changes to the `/heid*` skills,
+  narrative-heavy contracts, a standing "green-tests-prove-nothing" direction
-  not to this repo.
+  for the code-review gate, and regin's table-vs-signature consistency pass.
-- The booth set churns hard: 26 → 24 during this session as the sweeper ran.
+  The fourth is new and is the one with evidence behind it: a **contract-time
-  Re-count rather than trusting any number written here.
+  VACUITY PASS** — for each invariant, name a change that defeats it and check
+  the test goes red. Regin and Kimi proposed it independently on the U4
+  paraphrase round; the code-review panel then showed five of seven U4
+  falsifiers were vacuous, and heid rates that the strongest single data point
+  for it so far. See `persistent-memory.d/2026-09-22-vacuous-falsifiers.md`.
+- The booth set churns hard: 26 → 24 → 25 across the last two sessions as the
+  sweeper ran. Re-count rather than trusting any number written here.
 
 ## Recent decisions