Why Browser Tests Fail Everywhere

Short summary

Browser tests that pass locally but fail in CI are rarely random — CI exposes environmental differences your local workflow hides, such as production bundles, different feature flags, and headless timing. The fix is to reproduce the exact CI artifact locally, log environment differences before tests start, and wait for observable state transitions instead of arbitrary delays. Streaming and hydration add synchronization challenges that require asserting on specific evidence rather than timing.

• •CI failures usually expose real environmental differences, not randomness
• •Reproduce by building the exact CI artifact and serving it locally with matching env vars
• •Wait for observable state changes (notification IDs, counters, status rows) instead of arbitrary sleeps

Generated with AI, which can make mistakes.