shard(tick): 0934Z — corrective PRs merged; #3398 BSD-vs-GNU regression fix

docs/hygiene-history/ticks/2026/05/15/0934Z.md

@@ -0,0 +1,54 @@
+| 2026-05-15T09:34:00Z | claude-opus-4-7 | 596e842c | shard: PR #3395 + #3397 corrective fixes merged; PR #3398 BSD-vs-GNU sed regression fix; substrate-honest discipline regression observed | (PR #3399) | discipline-notes can regress during rewrites; need to preserve prior corrections explicitly |
+
+# Tick 0934Z — PR #3395 + #3397 merged; PR #3398 BSD-vs-GNU regression fix
+
+## Headline
+
+- **PR [#3395](https://github.com/Lucent-Financial-Group/Zeta/pull/3395)** (0911Z shard) MERGED → `ef4e1bb` on `origin/main`. Col 5 now correctly = `(PR #3395)`.
+- **PR [#3397](https://github.com/Lucent-Financial-Group/Zeta/pull/3397)** (0916Z corrective fix) MERGED → `3b53469`. Col 5 of 0916Z on main now correctly = `(PR #3396)`. Body col 4 restored to generic `(PR TBD)` discipline-name reference.
+- **PR [#3398](https://github.com/Lucent-Financial-Group/Zeta/pull/3398)** (0921Z shard) picked up 1 Codex P2 thread: BSD-vs-GNU sed portability — the same lesson Codex raised on PR #3393 earlier. I'd lost the portability note when rewriting the discipline in 0921Z. Fix commit `2ac9c7e` re-includes BSD form + GNU form + Perl portable form. Thread resolved.
+- Cron sentinel `596e842c` armed.
+
+## Δ since 0921Z
+
+| What | At 0921Z | At 0934Z |
+|---|---|---|
+| PR #3395 (0911Z fix) | OPEN with bb41604 | MERGED (`ef4e1bb`) |
+| PR #3397 (0916Z fix) | OPEN, auto-merge armed | MERGED (`3b53469`) |
+| PR #3398 (0921Z shard) | OPEN, just created | OPEN, 1 Codex P2 → fixed + resolved (`2ac9c7e`) |
+| Sed discipline portability | macOS-only re-form (regressed) | macOS + GNU + Perl (re-corrected) |
+| Cumulative session merges | 16 | 18 |
+
+## Substrate-honest observations
+
+### Discipline-notes can regress during rewrites
+
+PR #3393's Codex P2 (BSD-vs-GNU sed) was addressed at 0903Z with a portability note. When I rewrote the discipline in 0921Z to introduce the `@`-delimited pipe-anchored form, I FORGOT to carry forward the portability note. Codex correctly flagged it again on PR #3398.
+
+Pattern: when refining substrate (rewriting a discipline section), prior-acknowledged corrections can drop unless explicitly carried forward. Each refinement should diff against prior-merged versions to preserve learnings.
+
+Mitigation idea: before pushing a refined discipline section, grep prior shards in the same evolution-chain for review-thread feedback that has landed. The 0855Z shard had the BSD-vs-GNU note; the 0921Z rewrite should have preserved it. Future-Otto: review evolution-chain shards (0844Z → 0855Z → 0921Z → ...) when refining the discipline; check for prior reviewer-acknowledged corrections.
+
+### Corrective-PR strategy worked
+
+The corrective-cycle this session:
+
+| Discovery tick | Bug | Fix PR | Affected shards on main |
+|---|---|---|---|
+| 0921Z | line-1-anchored sed misfires on body (PR TBD) mentions | PR #3395 (in-flight 0911Z) + PR #3397 (0916Z on main) | 2 merged shards corrected |
+| 0934Z | discipline-rewrite lost BSD-vs-GNU portability note | PR #3398 commit `2ac9c7e` | (still in-flight; will land in 0921Z shard via #3398 merge) |
+
+Both fix paths used: (a) inline commit when PR still open (0911Z fixed before #3395 merged), (b) corrective PR when shard on main (0916Z via PR #3397). Both work; choice is dictated by merge-state.
+
+## Cron sentinel
+
+`596e842c` armed.
+
+## Next
+
+Cron-driven. Next tick:
+
+1. Verify PR #3398 + this tick's PR auto-merge fires
+2. Address any new threads
+3. PRs #3379 + #3381 (gate=UNKNOWN for many ticks) — investigate if they're stuck
+4. Consider canonicalizing the `@`-delimited + portability discipline in `.claude/rules/` (separate small PR)