flaky-test-detective

To find the root cause of flaky tests, classify the failure by separating ordering, shared state, timing, and infrastructure issues, then reproduce it deterministically with repeat runs and environment pinning until the pattern is measurable.

Tests that pass on retry but fail during parallel runs often suffer from shared state or ordering dependencies. Reproduce the parallel-only bug using shuffle seeds and parallel stress tests to isolate the specific timing or state collision.

Prove test stability by applying one-change-at-a-time remediation and comparing before-and-after run counts. Use repeated-run evidence to demonstrate the single demonstrated cause is resolved without adding retries or sleeps.

Yes, this approach handles quarantined tests and environment-dependent breakage by keeping quarantine decisions honest and time-bounded, while using environment checks to isolate and verify the specific conditions causing the intermittent failure.

Reproduce CI-only E2E test failures by pinning the environment, running repeated executions, and applying parallel stress to match CI conditions until the failure pattern becomes measurable and exposes the root cause.

Avoid retries or sleeps when fixing intermittent test failures if you cannot identify the root cause. Use evidence-based triage to classify the flake and prove stability with repeated-run counts before considering quarantine.