A restart, retry, or larger queue may clear an alert without preventing its return. Five Whys is a root-cause technique for tracing a failure beyond its first visible symptom: ask why it happened, use the answer to form the next question, and stop when you reach something you can change. Five is a prompt to keep investigating, not a required count. section 8.0 section 8.2.1
Intermittent uploads
Uploads fail because storage returns 503. Storage returns 503 because its request queue is full. Large files fill the queue faster than workers drain it because files are not size-validated before enqueue. Validation is absent because the original specification never set a maximum upload size. The countermeasure is a size limit at the API boundary, not simply a bigger queue. section 8.2.1
Back each answer with observed behavior, logs, metrics, traces, or a reproducible test. If an answer is a guess, gather evidence before asking the next why; an unsupported link can misdirect the entire chain. The questions expose assumptions, such as whether a full queue really means the queue is too small. section 8.2.1
Five Whys traces one cause chain, but it need not complete an investigation alone. Is/Is Not can narrow where a failure occurs; a fishbone diagram can check for causes the chain missed; Pareto analysis can help prioritize by impact. In post-incident reviews, people also need to feel safe describing what happened rather than defending their actions. section 8.2.5 section 13.2
Listing 8.4 AI approach to fixing a bug it was presented with
# AI-generated fix: retry until user appears
def test_user_exists():
user = create_test_user()
for attempt in range(5):
if db.find_user(user.id):
break
time.sleep(0.1)
assert db.find_user(user.id) is not NoneThis retry makes an intermittent database test pass without resolving its cause. Tracing why the assertion runs before the user exists leads to an asynchronous insert that the helper does not await; the durable fix is to await the write. section 8.4




