Technical proof: a recurring bug closed with a regression sensor

This case comes from a technical proof, not a client: it demonstrates the TriageBug flow applied to a bug that “had already been fixed” before — and kept coming back.

The reproducible scenario

A timezone calculation failed on dates near the day’s rollover, but only in some runs. The previous fix had been by guesswork: someone tinkered, the error disappeared from the screen and the hunt was declared over. Weeks later, it reappeared with a different face.

What triage did differently

Instead of repeating the guess, triage first made the bug deterministic: it pinned the input and the time conditions until it occurred reliably. With the reproducible case, the real cause was isolated — a timezone conversion applied twice — and the fix was minimal and targeted.

The step that closes the hunt

Before closing, a check was created that reproduced the original bug. After the fix, it passed; and it will fail again if the double conversion returns. It is that regression sensor that answers the question “is it really over?” without depending on whoever is on call.

The evidence that remains

The difference between the two fixes wasn’t the intelligence of whoever tinkered — it was the method. The second time, the end of the hunt is a repeatable result, not a fleeting relief.