chore(gates): prune stale fallow config, health baseline, and dead-code baseline

Merged
TH
4d ago
+22-429

Summary

Vitest 4.1.11 → 5.0.4. The local @vitest/coverage-v8 patch is gone: upstream now carries a module's script wrapper offset through the merge of worker coverage files instead of letting read order decide it (vitest-dev/vitest#10506, #10643), so the patch, its patchedDependencies entry and the cases pinning dist/wrapper-lengths.js delete. The converter defect is not fixed upstream — 5.0.4 pulls ast-v8-to-istanbul ^1.0.7, which still publishes a negative implicit-else count — so that patch is re-taken with pnpm patch. 16 files, 824 lines excluding the lockfile.

Reaching us: test.sequential removed (14 sites → { concurrent: false }); inline projects inherit root globalSetup/maxConcurrency; coverage.exclude matches without picomatch contains, so four file-kind patterns are root-free, holding the denominator at main's exact 1,730-file set. Node 22.12/24.13 and Vite 8.2.1 already satisfy v5.

recordEvent queues the event-log append, so open names events.ndjson before it exists; 4.1.11 drained the queue by timing, v5 doesn't. Fixed by awaiting the store's flushEvents.

Validation

9dc6119: check:affected --run, check:coverage-changed --base, check:coverage-crap --base (blocking) all exit 0. Unit lane, identical work (1,578 files / 13,438 tests): 158 s → 185 s. Coverage, alternating A/B under one lock: 248/265 s → 212/199 s. Lines 93.91 → 93.92, statements 92.14 → 92.15, functions 94.92 → 94.93, branches 85.42; 1,730 map files, zero impossible counts, no Unscored, no function ≥ 30 either side. mutation:run --modules daemon-ref-frame: 95.89 %, 70 killed / 0 survived (main 69 / 1).

Unresolved: install --frozen-lockfile fails ERR_PNPM_MINIMUM_RELEASE_AGE_VIOLATION — 5.0.4 published today 09:36–09:47Z, inside pnpm's 24 h default; self-clears ~2026-10-13T09:47Z. No exclusions added; local lanes used --config.minimum-release-age=0.