Rubber duck, but it asks questions

Stuck, and not yet sure what you're even stuck on.

I am stuck on a bug. Do not suggest fixes yet.

Ask me questions, one at a time, to narrow down where the problem is. After each answer, tell me what it rules in or out before asking the next.

Start with the broadest split you can: is this a problem of wrong data going in, wrong logic in the middle, or wrong handling of the result? Then keep halving.

Only once you're confident you've located it should you propose a cause. Then ask me to verify before suggesting a fix.

Here is what I know so far:

Chasing down a bug. Finding the cause when something fails, runs slow, or will not reproduce locally.