diagnosing-bugs

Non-deterministic bugs require raising the reproduction rate rather than finding a clean repro. Loop the trigger 100 times, parallelize execution, add stress, narrow timing windows, or inject sleeps until the failure rate is high enough to debug against.

Write a curl or HTTP script against the running dev server that asserts the exact symptom the user reported. The loop must be deterministic, fast, and able to go red on this specific bug, not just run without erroring.

Performance regressions require measurement before fixing: establish a baseline with a timing harness, performance.now(), a profiler, or query plans, then bisect to find the cause. Logs are usually the wrong tool for perf diagnosis.

Write the regression test before the fix, but only if a correct seam exists where the test exercises the real bug pattern as it occurs at the call site. If no correct seam exists, document that architectural gap instead of writing a shallow test.

Tagging every debug log with a unique prefix like [DEBUG-a4f2] makes cleanup a single grep operation at the end of the session. Untagged logs tend to survive and pollute the codebase, while tagged logs are reliably removed.

Stop and say so explicitly rather than hypothesizing without a signal. Ask the user for environment access, a redacted captured artifact like a HAR file or log dump, or permission to add temporary production instrumentation.