3.3Flaky Tests and Deterministic Design

A flaky test is a test that passes and fails without any code change. Flaky tests destroy trust in the test suite — when a failure might be "just flakiness," developers stop investigating failures.

Root Causes of Flakiness

Source	Example	Fix
Timing / race conditions	Asserting before async operation completes	Use event-driven waits, not sleep()
Shared state	Test A creates data that Test B depends on	Isolate test data per test
External dependencies	Test calls a live third-party API that is slow or down	Mock external services
Order dependency	Tests pass in sequence but fail when shuffled	Each test must set up its own preconditions
Non-deterministic data	Asserting on timestamps or random IDs	Assert on stable properties, use patterns/ranges

Anti-Pattern: Add retries to make flaky tests pass. The test still has a bug — you have just hidden it behind retries.

Pattern: Design tests for determinism from the start. When flakiness appears, investigate and fix the root cause. Track flaky test rates on a dashboard — a rising flaky rate is an early warning that test architecture needs attention.