What are the common causes of flaky tests in a backend codebase, and how do you fix them?

A flaky test is a test that passes and fails on the same code. Common causes:

• Shared state left by other tests: rows, cache keys, singletons, or mocks that were not reset.
• Order dependence: a test only passes if other tests run before it.
• Time: Date.now(), midnight, time zones, or real sleeps.
• Randomness without a seed.
• Network calls to real services.
• Race conditions: for example, the test checks the result before a job or event has finished.
• Resource clashes: fixed ports, or one database shared by parallel workers.

Fixes:

• Give each test its own data, and reset state in hooks like beforeEach and afterEach.
• Control time with fake timers or an injected clock.
• Seed randomness.
• Stub the network.
• Wait for a condition with a timeout instead of sleeping for a fixed time.
• Use random ports and a separate database per worker.

Treat flakiness as a bug. Quarantine the test, which means take it out of the main run, and give it an owner. Then fix the cause. Don't hide it behind endless retries.