# Flaky Tests and Stability Policy

Adopt an ownership-first reliability model.

## Principles

- each flaky test has an owner
- root cause first, no band-aid waits
- keep blocking suite healthy

## Quarantine policy (target)

- New/updated tests should prove stability before becoming blocking.
- If a test drops below acceptable pass rate, quarantine it and create follow-up.
- Reintroduce quarantined tests only after consecutive successful runs.

Suggested threshold model (aligned with Humand plan):

- warning below 95%
- quarantine below 90%
- return after 5 consecutive green runs

## Typical root causes

- unstable selectors
- async timing assumptions
- hidden data dependencies
- environment/test-order coupling

## Remediation checklist

1. stabilize selector strategy
2. remove implicit timing assumptions
3. make data setup explicit
4. verify isolation in parallel execution
5. rerun targeted scope multiple times