The Four Questions That Sort Any Failure

Ask them in order

1. Did it ever pass? Check git log on the test file. A test that was never green is a new-test problem, not a regression.
2. Does it fail locally? CI-only failures are environment problems — lesson 5 covers those. Failing on your machine is good news: you can see it.
3. Does it fail alone? Run just that one test. Passes alone but fails in the suite? Another test is leaking state into it.
4. Does it fail every time? Run it ten times. 10/10 is a real bug — follow the trace. 3/10 is a race — timing is the suspect.

# 1. did it ever pass?
git log --oneline -- tests/test_checkout.py

# 2 + 3. does it fail locally — alone, and in company?
pytest tests/test_checkout.py::test_guest_checkout
pytest tests/

# 4. does it fail every time?
for i in {1..10}; do pytest -q tests/test_checkout.py::test_guest_checkout; done

What the answers mean

• Never passed → fix the test's assumptions.
• Fails locally, alone, every time → an ordinary bug. Read the trace (next lesson).
• Fails only in the suite → state leak. Find the test that runs before it.
• Fails sometimes → a race. Measure it (lesson 4).

Grounded in the pytest docs on selecting and re-running tests