What causes flaky tests and how do you fix them?

Deep dive

Common sources:

• Race conditions, async not awaited.
• Shared state between tests.
• Reliance on external services or clocks.
• Order-dependent tests.

Fix strategies:

• Use deterministic clocks and random seeds.
• Reset state between tests.
• Replace sleeps with condition-based waits.
• Quarantine flaky tests and track trends.

Examples

Condition-based wait:

waitUntil(() => element.isVisible())

Common pitfalls

• Retrying flaky tests without fixing root cause.
• Ignoring flakes in CI until they pile up.
• Using real network dependencies in unit tests.

Interview follow-ups

• How do you detect flaky tests early?
• When is retry acceptable?
• How do you keep E2E tests stable?