[blafourcade]

🎯 What & why

The telemetry mutation run took 45 to 62 min locally. Each mutant ran the slow integration test files before the unit test that would have killed it in milliseconds. After this change the run takes 6.7 min locally (CI shards: 3.6, 4.9 and 6.9 min). No test was removed or weakened.

🛠️ How it works

• Unit tests run first, smallest file first, for mutation only. A sequencer (mutation-sequencer.ts) is loaded only by vitest.mutation.config.ts. Stryker stops a mutant at its first failing test, so the cheapest test that kills it now runs first. The sequencer changes the order of tests, never which tests run. pnpm test keeps its order. Measured: 45.5 to 8.9 min.
• Fewer static mutants (330 to 158). Values that tests built at module load are now built inside the tests that use them. The assertions and it.each titles are unchanged. Test runs per full telemetry run drop from 94k to 39k.
• timeoutFactor 3 (stryker.conf.json:9). 165 of the old timeouts were slow kills, not hangs, and are now counted as killed. timeoutMS stays at 20000, the value test(cli): count a mutant as killed only when a test failed #844 set.
• pnpm test. The 33 s telemetry report e2e file is split into 3 files, with 22 tests before and after. The golden tests capture each cell twice instead of three times; the determinism check (capture 2 against capture 1) is kept. The shared code lives in two-captures.ts. The overall wall-time gain is within noise; the e2e subset went from 17.1 s to 12.1 s.

🧪 How to verify

• cd cli && pnpm test: 6641 passed, 1 skipped.
• pnpm test:mutation:telemetry, compared mutant by mutant with the baseline:
  • no mutant killed in the baseline survives;
  • 165 went from Timeout to Killed;
  • 7 went from Timeout to Survived, and they already survived in a fresh baseline run on unchanged code;
  • 2 went from Killed to Timeout, and these are real hangs in ledger-lock.ts.
• Every mutation scope holds its floor. framework is at 93.0 against 93.1 in CI before this change, with no regression per mutant.
• Sequencer guard: mutation-sequencer.unit.test.ts goes red when sort returns the files unsorted, or when a file size reads 0.
• Timings, the per-mutant comparison and the load on each run are in evidence.md. Independent review, 2 rounds: review.md.

⚠️ Heads-up

• Telemetry score: 96.35, against 96.52 before (floor 96). The baseline counted 7 survivors as timeouts because the machine was loaded. Counted the same way, both runs score 96.33 on the 4169 mutants they share. The maintainer accepted this score rule; it is recorded in the contract.
• This PR touches the shared mutation harness, so CI runs every mutation scope.
• git-repository-locator-adapter.ts:78 is a flaky mutant: Timeout, Survived or Killed depending on the run. It is reported, not fixed.
• The merge commit 7639a681 has git's default message. The squash replaces it.

🔗 Linked issue

Closes #1036

✅ I certify

• I DO CERTIFY I READ EACH LINE OF THE PULL REQUEST BECAUSE I AM A SOFTWARE ENGINEER, NOT A AI PUPPY.

🤖 Generated with Claude Code

https://claude.ai/code/session_01VWNxk63AGKkqE8HRqHLjGp