CI Reliability Plan for a Slow, Flaky Suite
Cut wall-clock time and flake rate of a CI test suite without reducing coverage
A staged plan with sharding and selection changes, retry and quarantine policy, diagnostic artefact configuration, and measurable targets per stage.
Role
You are the CI owner for {{suite_name}}.
Task
Produce a staged plan that reduces suite duration and flake rate while preserving coverage, covering parallelisation, retry and quarantine policy, and the diagnostics captured on failure.
Context
{{suite_name}} takes {{current_duration}} on {{runner_profile}} with flake rate {{flake_rate}}, current retry policy {{retry_policy}} and diagnostics {{failure_diagnostics}}.
Inputs
- {{current_duration}}
- {{flake_rate}}
- {{runner_profile}}
Constraints
- State a measurable target and a measurement method for each stage
- Never let retries hide a genuine failure without recording it
- Capture failure diagnostics selectively so cost stays bounded
- Do not reduce coverage to hit a duration target
Output Format
Markdown: staged plan table (stage, change, expected effect, measurement), retry and quarantine policy, diagnostics configuration.
Quality Criteria
- Every stage has a measurable target
- Retry policy preserves failure visibility
- Quarantine has an exit condition
- Diagnostics cost is bounded