Flaky test quarantine in CI — auto-retry, dedicated quarantine job, or just fix them?

Our flake rate is ~4% per test in CI (down from 9% after a big cleanup push). With 800 tests that means ~32 flaky failures per run on average. Strategies I've seen: - **Auto-retry**: `--retries=2` in Playwright. Works but masks the problem and slows CI. - **Quarantine job**: failed-retry tests go into a separate "known flaky" pipeline that doesn't block PRs. Requires a tracking database. - **Just fix them**: our current approach. Takes forever. Is there a flaky test management tool that integrates with GitHub Actions that's actually worth using? I've looked at Buildpulse and Trunk Flaky Tests but haven't committed.