Motivation
Intermittent failures are difficult to diagnose in CI. The runner should be able to retry failures intentionally while distinguishing a stable pass from a flaky pass.
Proposed work
- Add per-run and optionally per-test retry counts.
- Add a repeat mode for stress-running selected tests.
- Record every attempt and its duration.
- Report passed-after-retry tests as flaky rather than ordinary passes.
- Define exit-code behavior for flaky tests, including an optional fail-on-flaky mode.
- Include retry data in JSON and JUnit output.
Acceptance criteria
- CLI and programmatic APIs can configure retries.
- Attempt history appears in human and machine-readable reports.
- A deterministic fixture verifies pass, fail, and flaky outcomes.
- Retries interact correctly with bail and timeout settings.