Flaky Test Hunter

You are running the "Flaky Test Hunter" loop.

Settings (edit before running):
- Test command: npm test
- Repeat runs to detect flakes: 20
- Clean streak required: 30 consecutive passes
- Max iterations: 6

Goal: find intermittently failing tests, fix their root causes, and prove each fix with a clean streak.

Each iteration:
1. Run the test command repeatedly (use the runner's repeat or shuffle options if it has them). Record every test that fails at least once and how often.
2. Take the most frequent flaky test. Read it and the code it covers, and classify the cause: timing or sleeps, shared or leaked state, order dependence, real clock or timezone, unseeded randomness, or real network calls.
3. Fix the root cause. Prefer explicit waits, per-test setup and teardown, fake timers, fixed seeds, and local fakes for network calls.
4. Run that test file for the clean streak count. If it fails even once, the fix is not done.
5. Log: test name, failure rate before, cause, fix, streak result.

Stop when:
- A full detection pass finds no flaky tests. Success.
- You reach the max iteration count.

Rules:
- Never add retries, increase timeouts as the only fix, or mark tests as skipped.
- Do not delete assertions or loosen them to make a test pass.
- If the flake is a real race in production code, fix it there and explain the bug.
- Keep each fix limited to the flaky test and the code behind it.

Final report: each flaky test, its cause category, the fix, and its before and after failure rate.