chore: dead-code sweep 2026-07-27 — drop unused export modifier on SubscribeSourceFactory

[mbret]

Automated strict dead-code sweep (knip over library + dev entries, plus a production-entries-only pass, every candidate verified by repo-wide grep).

Removed

• export modifier on SubscribeSourceFactory (src/lib/binding/useSubscribe/types.ts) — the type itself is live (used by SubscribeSource in the same file), but the export was dead surface. Evidence: repo-wide grep for SubscribeSourceFactory matches only its own declaration and the same-file SubscribeSource reference; src/lib/binding/useSubscribe/index.ts re-exports only ./useSubscribe, not ./types, so the type never reaches the package's public API via src/index.ts.

Candidates investigated and kept

• src/lib/state/persistence/adapters/MockAdapter.ts — flagged unused in production-only analysis, but it is live test infrastructure imported by persistSignals.test.ts.
• src/lib/utils/filterObjectByKey.ts — same situation: imported by useObserve.selector.test.tsx as a test helper.

No unused files, no unused package dependencies found. Public API exported through src/index.ts was treated as off-limits (published npm package).

Gates

Baseline on clean main and after the change, both fully green:

• npm run check (biome) ✓
• npm run build (tsc + vite) ✓
• npm run test:ci — 23 files, 134 tests ✓

🤖 Generated with Claude Code

https://claude.ai/code/session_01R9ch8GmdfsE2RPy8adBGPo

---

Generated by Claude Code

[mbret]

CI is red on Check, but the failure is a pre-existing flaky test, not this PR: useQuery$.test.tsx > should return consecutive results (the interval(5) variant at line 74) intermittently sees emissions 3 and 4 coalesce into a single render, so states[4] gets data: 4 instead of data: 3.

Evidence it's unrelated to this diff:

• This PR only removes an export modifier on a type — erased at compile time, no runtime effect, and useQuery$ is untouched.
• The failure reproduces on a clean origin/main checkout: 1 failure in 4 full-suite runs locally, with the identical assertion diff.
• On this branch the suite passes most runs too (flaked 1 in 3 locally, and the pre-push run was green).

I've re-run the failed job; given the observed ~25–30% failure rate it should go green within a retry or two. The test's own comment ("interval big enough so react does not skip some render") hints at the hazard — it may be worth deflaking separately, but that's out of scope for a dead-code sweep.

---

Generated by Claude Code