shard(tick): 1731Z — parallel work caught Vera+Riven MEMORY.md off-by-one PR citations

docs/hygiene-history/ticks/2026/05/15/1731Z.md

@@ -0,0 +1,38 @@
+# Tick 1731Z — parallel speculative work caught Vera + Riven MEMORY.md off-by-one PR citations
+
+## Headline
+
+- [PR #3525](https://github.com/Lucent-Financial-Group/Zeta/pull/3525) (1718Z shard) wait-CI; auto-merge armed.
+- **Self-caught bug**: while waiting on #3525 CI, audited my own recent MEMORY.md files (Vera, Riven) and found 2 off-by-one PR citations. Filed [PR #3526](https://github.com/Lucent-Financial-Group/Zeta/pull/3526) — 2-line fix, auto-merge armed (MERGEABLE).
+- Cron sentinel `<<autonomous-loop>>` confirmed live from prior tick.
+
+## Bug detail
+
+| File | Cited | Should be | Why wrong |
+|---|---|---|---|
+| `memory/persona/vera/MEMORY.md` | PR #3515 | PR #3516 | #3515 is the 1524z fix-PR |
+| `memory/persona/riven/MEMORY.md` | PR #3514 | PR #3513 | #3514 is the Alexa migration |
+
+Both verified via `gh pr view` title cross-check. Same root cause — authoring under bandwidth pressure with adjacent PR numbers in flight; "the migration PR" got off-by-one twice.
+
+## Per-tick discipline trace (canonical 7-step)
+
+1. **Refresh**: `gh pr view 3525` showed wait-CI (5 IN_PROGRESS, 16 SUCCESS). Real named dependency with bounded ETA (~few minutes).
+2. **Holding-discipline**: named dependency present, but per [`.claude/rules/holding-without-named-dependency-is-standing-by-failure.md`](../../../../../../.claude/rules/holding-without-named-dependency-is-standing-by-failure.md) NOT a reason to idle — pick parallel work.
+3. **Pick work**: tier-1 audit of my own recent landings → caught real bug.
+4. **Verify**: `gh pr view` title cross-check on PRs #3513–#3516 verified the correct mapping.
+5. **Shard**: this file.
+6. **CronList**: sentinel confirmed live from prior tick.
+7. **Visibility**: PR #3526 + this shard.
+
+## Substrate-honest meta-note
+
+The 1718Z tick *surfaced* the §33 cascade residual; this tick *caught the bug in my own §33 work*. Both are the same parallel-work-while-CI-runs discipline. The discipline produces output even when waiting on dependency — that's the point.
+
+The off-by-one error class is a small instance of the broader "adjacent PR numbers under bandwidth pressure" pattern. Could be caught by a future static lint that cross-checks PR citations in MEMORY.md / persona docs against actual PR titles via `gh api`. Filing as a thought, not yet a B-NNNN — would compose with [B-0532](../../../../../backlog/P3/B-0532-backlog-graph-consistency-lint-parent-child-status-mismatch-2026-05-15.md) (parent-child status-mismatch lint).
+
+## Composes with
+
+- [`.claude/rules/holding-without-named-dependency-is-standing-by-failure.md`](../../../../../../.claude/rules/holding-without-named-dependency-is-standing-by-failure.md) — parallel work even when there IS a named dependency
+- [`.claude/rules/refresh-before-decide.md`](../../../../../../.claude/rules/refresh-before-decide.md) — `gh pr view` title cross-check was the falsifier
+- [`.claude/rules/substrate-or-it-didnt-happen.md`](../../../../../../.claude/rules/substrate-or-it-didnt-happen.md) — fix landed as durable substrate (PR), not just a "I noticed this" mention