A production rollout exposes a contract mismatch between the UI boundary and the owning service. The affected area is feature flags.
Repair the feature flags tests implementation so it classifies stable, flaky, and consistently failing tests from repeated independent runs.
Evidence
- Affected surface: Full-stack frontend focus / feature flags.
- Observed failure family: triage an unreliable test suite.
- Scope policy: zoneState must equal healthy.