shard(tick): 0838Z — PR #3388 merged; 2nd stale-Codex-review pattern on PR #3389; mechanization candidate filed

docs/hygiene-history/ticks/2026/05/15/0838Z.md

@@ -0,0 +1,73 @@
+| 2026-05-15T08:38:00Z | claude-opus-4-7 | 596e842c | shard: PR #3388 merged; PR #3389 stale-Codex-review pattern recurred (2nd this session); restraint discipline maintained on mechanization | (PR #3391) | (PR TBD)+post-create-fix sequence is structurally vulnerable to stale reviews; mechanization candidate not yet justified given low per-occurrence cost |
+
+# Tick 0838Z — PR #3388 merged; 2nd stale-Codex-review pattern on PR #3389; restraint
+
+## Headline
+
+- **PR [#3388](https://github.com/Lucent-Financial-Group/Zeta/pull/3388)** (0820Z shard) MERGED → `250cbae` on `origin/main`.
+- **PR [#3389](https://github.com/Lucent-Financial-Group/Zeta/pull/3389)** (0829Z shard) picked up a stale-Codex-review thread: Codex P2 flagged `(PR TBD)` placeholder but commit `6dd0fb0` had already replaced it with `(PR #3389)`. Same pattern as PR #3388's stale-Copilot thread one tick ago. Reply + resolve.
+- **5 in-flight PRs remain** (#3379, #3381, #3382 (bulk-fix), #3389, this tick's #3391). All in CI wait.
+- Cron sentinel `596e842c` armed.
+
+## Δ since 0829Z
+
+| What | At 0829Z | At 0838Z |
+|---|---|---|
+| PR #3388 (0820Z shard) | OPEN, 2 threads resolved | MERGED (`250cbae`) |
+| PR #3389 (0829Z shard) | OPEN, wait-ci | OPEN, 1 stale Codex P2 → reply + resolved |
+| Stale-review occurrences this session | 1 (on PR #3388) | 2 (added one on PR #3389) |
+| Cumulative session merges | 10 | 11 |
+
+## Substrate-honest observations
+
+### Stale-review pattern is structural, not accidental
+
+Two stale-Copilot/Codex `(PR TBD)` reviews this session. Both produced because:
+
+1. I open the PR via `gh pr create` with `(PR TBD)` in the pipe-row
+2. Copilot/Codex auto-review fires within seconds against the initial commit
+3. I push the post-create-fix replacing `(PR TBD)` with `(PR #NNNN)` ~10s later
+4. The review arrives in the GitHub UI a minute later, but it's pinned to the pre-fix commit
+5. Thread reads "still says `(PR TBD)`" — which is true at the snapshot but false at HEAD
+
+The window between (1) and (3) is the structural vulnerability. The manual sequence can't close it; the reviewer snapshot is faster than the human/Otto sed+commit+push cycle.
+
+### Mechanization candidate — first proposal does NOT eliminate the race
+
+An initial mechanization proposal was:
+
+1. Write shard with `(PR TBD)`
+2. `git commit` (local, no push)
+3. `git push` (creates branch on remote)
+4. `gh pr create` (opens PR)
+5. Capture new PR# from gh output
+6. `sed -i '' '1s/(PR TBD)/(PR #NNNN)/' <file>` on the shard
+7. `git commit --amend --no-edit` (rewrite the SOLE commit)
+8. `git push --force-with-lease` (replaces remote ref atomically)
+
+**This does NOT eliminate the stale-review race** (per Copilot review on PR #3391 thread 4). The window between (4) and (8) still exists: `gh pr create` triggers Copilot's review-snapshot of the initial commit before the force-push at (8) replaces it. Reduces the window but doesn't close it. Result: snapshot may still see `(PR TBD)`.
+
+**Real fix options** (more substantial):
+
+- **Option A — no PR# in pipe-row**: drop the PR-reference column from the pipe-row schema entirely. Traceability via PR title + branch name + tick-history-index file (generated by a CI post-merge step). Eliminates the `(PR TBD)` placeholder need entirely. Requires schema-validator update.
+- **Option B — index-file traceability**: keep pipe-row PR# field but populate via a separate index file (`docs/hygiene-history/ticks-pr-index.json`) maintained by a CI post-merge step. Pipe-row uses a placeholder that's accepted as final state. No post-create-fix needed.
+- **Option C — accept the stale-review cost**: the manual sequence works; per-occurrence cost is ~2 tool calls; observed frequency ~50% of PRs; per-session cost 2-3 tool calls. Cheaper than building either Option A or B.
+
+**Decision deferred**: substrate-honest acknowledgment that the simple mechanization doesn't solve the problem. Option C (status quo) is the operational answer until/unless someone designs Option A or B. Filing this as a follow-on candidate rather than shipping a partial mechanization that misrepresents the race-closure.
+
+### Restraint discipline maintained
+
+This tick had two carry-forward candidates: claude-loop investigation, `(PR TBD)` mechanization. Did NOT start either. The 4 in-flight PRs ARE the named-dependency wait. Address what surfaces; don't manufacture new substrate-threads.
+
+## Cron sentinel
+
+`596e842c` armed.
+
+## Next
+
+Cron-driven. Next tick:
+
+1. Verify PR #3379 + #3381 + #3382 + #3389 + this tick's PR auto-merge fires
+2. Address any new review threads as they surface
+3. Watch PR #3323 for Lior response — Lior loop still active per archive-automation branches landing each fetch
+4. If quiet next tick + tick budget permits, pick claude-loop investigation (the last unknown in B-0530's prune-race candidate list)