Flaky test bingo: 12 squares I have seen in real life (with fixes)

Hello everyone) I collect flaky test causes like other people collect stamps. Here is my bingo card from the last 4 years. All real, all happened to me or my team. If you get a full row in one sprint, you deserve a day off.

Row 1: Time

• 1. Test uses real current time. Fails at midnight, at end of month, on DST change. Fix: freeze time. In Playwright page.clock.setFixedTime(), in Jest or Vitest fake timers with a fixed system time.
• 2. Timezone of CI is not timezone of laptop. CI is UTC, my laptop is Kyiv, so the date is different for 3 hours every night. Fix: set timezoneId in the Playwright config and the TZ environment variable for unit tests.
• 3. Hardcoded wait. waitForTimeout(3000) works on your machine, CI is 2x slower. Fix: wait for a condition, not for time. Web-first assertions like await expect(locator).toBeVisible() retry by themselves.

Row 2: Data

• 4. Shared test user. Somebody changed the password, language or role of "qa_main". Fix: every test creates its own data through the API in a fixture.
• 5. Tests depend on order. Test B passes only because test A created something before it. Run in a different order or shard the suite and it dies. Fix: every test sets up its own state. Playwright runs files in parallel, so this shows up quickly.
• 6. Unique field collision. Two parallel workers create a user with email test@test.com. Fix: put the worker index and a timestamp into the email, for example 'user-' + testInfo.workerIndex + '-' + Date.now() + '@example.test'.

Row 3: The page

• 7. Click before the element is ready. Button is visible but disabled while JS loads, or covered by the cookie banner. Fix: Playwright waits for actionability by itself, but it will not close your banner. Close it in a fixture.
• 8. Animation. Element is still moving when you click. Fix: reducedMotion: 'reduce' in the config, and make your app respect prefers-reduced-motion (real users will thank you too).
• 9. Locator matches 2 elements. Works while one of them is hidden, fails when both render. Fix: locate by role and name, getByRole('button', { name: 'Save' }). Strict mode will tell you when it is ambiguous.

Row 4: Environment

• 10. Third-party service inside the test. Payment sandbox is down every second Thursday. Fix: mock it with page.route() in UI tests and test the real integration separately, with a few tests only.
• 11. Resource limits. CI runner has 2 CPU and 4 GB, your laptop has 10 cores. Too many workers, everything is slow, timeouts everywhere. Fix: set workers for CI explicitly.
• 12. Free square: "it passed on rerun". This is not a fix. This is a lottery ticket)

How I keep it under control

• Retries only in CI (retries: process.env.CI ? 2 : 0), and every test that passed on retry goes to a report. Passed on retry is still a failure, just quieter.
• Quarantine list with owner and date. No owner, no quarantine.
• Trace on first retry, so when it fails you have DOM, network and console of the exact moment.

Write in comments which square you have this week) I have number 4. Again.