hygiene(#268): pr-preservation drain-log for #435 (live-lock cadence reword)

docs/pr-preservation/435-drain-log.md

@@ -0,0 +1,118 @@
+# PR #435 drain log — drain follow-up to #148: cadenced live-lock + grammar
+
+PR: <https://github.com/Lucent-Financial-Group/Zeta/pull/435>
+Branch: `drain/148-followup-cadenced-livelock-and-grammar`
+Drain session: 2026-04-25 (Otto, post-summary continuation autonomous-loop)
+Total threads drained: 3 across 2 waves (2 + 1 cascade)
+Rebase context: clean rebase onto `origin/main`; no conflicts.
+
+Per Otto-250 (PR review comments + responses + resolutions are
+high-quality training signals): full per-thread record with reviewer
+authorship, severity, outcome class.
+
+This PR was a drain follow-up to #148 (`docs/plans/why-the-factory-
+is-different.md`) catching reviewer findings on the live-lock-smell
+audit cadence claim and a heading grammar issue. The wave-2 cascade
+caught a deeper issue: the doc was implying a multi-BACKLOG-item
+state when only one BACKLOG row exists for live-lock cadence work.
+
+---
+
+## Wave 1 (initial drain — 2 threads)
+
+### Thread 1.1 — `:111` — Cadence row in FACTORY-HYGIENE.md doesn't exist (Codex P2)
+
+- Reviewer: chatgpt-codex-connector
+- Thread ID: `PRRT_kwDOSF9kNM59kGrg`
+- Severity: P2
+- Finding: text said operators can run the live-lock audit "on the
+  cadence row in `docs/FACTORY-HYGIENE.md`" but that file has no row
+  for `tools/audit/live-lock-audit.sh`; `docs/BACKLOG.md` still
+  tracks round-close wiring as unfinished.
+- Outcome: **FIX (option b — reword to current-truth)** — reworded
+  to factually describe current state: "audit is currently run on
+  demand by operators. Promoting it to a cadenced row in
+  `docs/FACTORY-HYGIENE.md` and wiring per-commit CI / hook
+  integration are separate BACKLOG items." Adopted reword option
+  over option a (adding the hygiene row) because the audit's
+  per-commit wiring is the substantive blocker the BACKLOG tracks;
+  the doc shouldn't claim cadence already shipped. Commit `2615fb4`.
+
+### Thread 1.2 — `:112` — Same finding (Copilot P1)
+
+- Reviewer: copilot-pull-request-reviewer
+- Thread ID: `PRRT_kwDOSF9kNM59kG0t`
+- Severity: P1
+- Outcome: **FIX (combined with 1.1)** — same fix as Thread 1.1;
+  one rephrase resolved both findings. Cross-reviewer convergence
+  (Codex P2 + Copilot P1) raised confidence the issue was real.
+
+## Wave 2 (post-#1 cascade — 1 thread)
+
+### Thread 2.1 — `:113` — Implied multiple BACKLOG items vs actual single row (Codex P2)
+
+- Reviewer: chatgpt-codex-connector
+- Thread ID: `PRRT_kwDOSF9kNM59kKLD`
+- Severity: P2
+- Finding: my prior reword said "Promoting it to a cadenced row ...
+  AND wiring per-commit CI / hook integration are separate BACKLOG
+  items" (plural). But the only live-lock backlog entry is
+  `docs/BACKLOG.md` Live-lock-smell cadence row (currently around L1452 in the P1 tooling section; line numbers drift, so the stable identifier is the heading 'Live-lock smell cadence (round 44 auto-loop-46 absorb, landed as `tools/audit/live-lock-audit.sh` + hygiene-history log)'), which tracks a single row with
+  follow-ups around round-close wiring, threshold tuning, and
+  PR-in-flight class. One row, multiple sub-items — not "separate
+  BACKLOG items."
+- Outcome: **FIX** — reworded to point at the actual single
+  existing BACKLOG row (`docs/BACKLOG.md` Live-lock-smell cadence
+  heading; was L1313-1328 at drain time, has since drifted to ~L1452
+  per the stable-identifier-vs-line-number-xref pattern) and name
+  its
+  existing sub-items: round-close cadence wiring, threshold tuning,
+  PR-in-flight class. Promoting to a cadenced
+  `docs/FACTORY-HYGIENE.md` row composes with that same row rather
+  than implying a separate BACKLOG entry. Doc is now verifiable
+  against current BACKLOG state. Commit `6171ec0`.
+
+---
+
+## Pattern observations (Otto-250 training-signal class)
+
+1. **Cross-reviewer convergence on Wave 1 raised quality signal.**
+   Threads 1.1 (Codex P2) + 1.2 (Copilot P1) flagged the same
+   missing-FACTORY-HYGIENE-row issue. Same shape as #432's
+   `warn` unbound (Codex P1 + Copilot P0). When two unrelated
+   reviewers converge with high confidence, the prior on "real
+   bug" goes way up.
+
+2. **My-fix-on-wave-1-introduced-the-wave-2-finding.** This is a
+   clean instance of a self-induced cascade: my reword (Wave 1)
+   said "separate BACKLOG items" (implying plural), Codex (Wave 2)
+   caught that the actual BACKLOG state has one row with multiple
+   sub-items. Pattern: when fixing a claim, verify the new claim
+   is also accurate — don't just remove the bad claim, replace it
+   with a verified-against-current-state claim.
+
+3. **"Reword option (a) vs (b)" decision template.** Wave 1's
+   reviewer suggestion offered two options: (a) add the hygiene
+   row (mechanism-first), (b) reword to current truth (description-
+   first). Option (b) was correct here because the substantive
+   blocker (per-commit wiring) is the tracked BACKLOG work; the
+   doc should match work-in-progress reality rather than claiming
+   shipped state. The (a)-vs-(b) framing generalizes: when a doc
+   asserts X but X doesn't exist, prefer reword-to-current-truth
+   over add-the-thing-asserted unless the thing is small + isolated.
+
+4. **PR-mechanics observation: 4 of the 7 PRs drained at u=0+
+   reviewer-cascade in this session went through this same wave-1
+   + wave-2 pattern.** #135, #231, #432, #435 all had post-merge
+   cascade waves catching what wave-1 fixes exposed or what new
+   reviewer-state revealed. The reviewer-cascade is a consistent
+   property of the merge-trigger surface, not a per-PR oddity.
+
+## Final resolution
+
+All 3 threads resolved across 2 waves (`2615fb4` + `6171ec0`).
+PR auto-merge SQUASH armed; CI cleared; PR merged to main as
+`a5e3eff`.
+
+Drained by: Otto, post-summary autonomous-loop continuation, cron
+heartbeat `f38fa487` (`* * * * *`).