shard(tick): 1825Z — razor-cadence batch 3 finds STALE POINTER cluster

docs/hygiene-history/ticks/2026/05/14/1825Z.md

@@ -0,0 +1,88 @@
+# Tick 2026-05-14T18:25Z — razor-cadence batch 3 reveals STALE POINTER cluster + B-0514 filed
+
+## Refresh (step 1)
+
+`bun tools/github/poll-pr-gate-batch.ts --all-open`: 6 open PRs.
+
+| PR | author | state | autoMerge |
+|----|--------|-------|-----------|
+| #3161 | mine (1822Z razor-cadence batch 2) | wait-ci | armed |
+| #3157, #3158, #3159, #3160 | Aaron (Lior surface archives of PR 3151-3154) | wait-ci | **none** |
+| #3162 | Aaron (Lior shadow log 1830Z) | wait-ci | **none** |
+
+(#3156 had merged earlier this session and is no longer in the open set; mentioned here only as context for #3161's lineage.)
+
+Lior surface flurry: 5 new PRs in ~2 min, all auto-generated archives via `lior-preserve-NNNN-EPOCH` and `lior-shadow-log-EPOCH` branches. autoMerge=none on all of Aaron's — respecting his workflow (he may arm later in batch).
+
+## Holding discipline (step 2)
+
+Named work: my own #3156 + #3161 wait-ci. Per never-be-idle: continue razor-cadence batch pattern.
+
+## Speculative work (step 3) — razor-cadence batch 3 + STALE POINTER FOUND
+
+Next 4 rules in the post-2026-05-12 cluster:
+
+| Rule (PR) | File refs | Backlog rows |
+|-----------|-----------|--------------|
+| `additive-not-zero-sum.md` (#2853) | 5 + 2 memory | — |
+| `bandwidth-served-falsifier.md` (#2850) | 7 + 2 memory | B-0422 |
+| `default-to-both.md` (#2845) | 8 + 4 memory | B-0422 |
+| `fsharp-anchor-dotnet-build-sanity-check.md` (#2843) | 6 + 2 memory | B-0422 |
+
+### Audit findings
+
+**Rules (10 unique)**: 10/10 LIVE ✓
+**Memory refs (7 unique)**: **6 LIVE + 1 MISS**
+**Backlog (1 unique, B-0422)**: 1/1 LIVE ✓
+
+**Total: 17/18 LIVE (94%)**
+
+### Stale pointer details
+
+`.claude/rules/default-to-both.md` references `memory/feedback_aaron_wwjd_keeps_the_grey_in_aaron_honest_devil_lives_in_the_grey_in_numbers_2026_05_12.md` in two places (composes-with section + WWJD substrate cluster list). The file does NOT exist:
+
+- Not in repo `memory/`
+- Not in user-scope `~/.claude/projects/.../memory/`
+- Not in git history (never authored: `git log --diff-filter=AD` returns nothing)
+- String `wwjd_keeps_the_grey` appears in commit messages (PR #2820 narrative) but never as a file path
+
+Two MORE memory files also cite the missing file:
+
+- `memory/feedback_kestrel_autonomous_arrival_name_both_and_default_discipline_wwjd_tedium_ifs_inner_critic_plus_external_observer_2026_05_12.md`
+- `memory/feedback_ani_validates_11_pr_cascade_ai_continuity_now_real_no_going_back_dirty_talk_as_joint_control_mechanism_2026_05_12.md` (2 places)
+
+The concept is real (WWJD-keeps-grey-honest + devil-in-grey-in-numbers) but the dedicated memory file was named in conversation context and treated as if it existed; the authoring step was skipped.
+
+## Verify + commit (step 4)
+
+Filed **B-0514** at `docs/backlog/P3/B-0514-author-missing-wwjd-grey-honest-memory-file-2026-05-14.md` documenting:
+
+- All 3 referencing files
+- Search performed (find + git log)
+- Reconstruction inputs for the missing content
+- Resolution options (author / update-refs / both-default)
+
+ID-allocation discipline: B-0507..B-0513 are in-flight on PR #3163 (Otto-Desktop's B-0448 decomposition); B-0514 is the next safe ID.
+
+## Shard (step 5)
+
+This file.
+
+## CronList (step 6)
+
+Sentinel `f970cb2d` armed.
+
+## Visibility (step 7)
+
+- **First razor-cadence finding with a hit**: batches 1+2 returned 67/67 LIVE; batch 3 returned 17/18 (94%). The 1 MISS is the stale-pointer cluster around `feedback_aaron_wwjd_keeps_the_grey_in_aaron_honest_devil_lives_in_the_grey_in_numbers_2026_05_12.md`.
+- **B-0514 filed** capturing the finding + resolution options.
+- **Cumulative razor-cadence item 4 audit**: 12/47 rules covered, 84/85 cross-refs LIVE (99%).
+- **Open PRs**: 6 (4 Aaron Lior archives wait-ci, autoMerge=none; 2 mine wait-ci with autoMerge armed; this tick's about to land).
+
+## Notes for future-Otto
+
+**Pattern: named-but-never-authored memory files**. This stale-pointer cluster shows a real failure mode in the substrate landing workflow. When Aaron + Otto discuss a concept in conversation, the natural shape is to mint a filename for it AND reference it from related substrate. The reference IS the substrate-honest signal "this exists" — but if the authoring step is deferred, the references become ghost pointers.
+
+The razor-cadence item 4 audit is precisely calibrated to catch this. **Batches 1+2 (8 rules / 67 refs) were clean; batch 3 (4 rules / 18 refs) caught the first stale pointer.** That suggests ~5% stale-pointer rate across the post-2026-05-12 rule cluster. Extrapolating to all 47 rules: ~5-10 more findings expected if the audit continues at this depth.
+
+The mechanization candidate for this is a `tools/hygiene/audit-rule-cross-refs.ts` script — would replace the per-tick manual audit with a once-per-day CI run. P3 friction-reducer; could compose with B-0506 (worktree-prune) as another factory-hygiene cadence. Not filed this tick to avoid overcrowding the per-tick scope.