Deleting a personal API token reported a successful removal whether or not there had been a token to remove. Click Delete twice, or aim at a token a session on another device had already revoked, and both attempts came back the same way: done.

Nothing was lost by it — the token really is gone in every one of those cases, which is why it went unnoticed. What was lost was the ability to tell the two apart. A tidy-up that expected to clear a handful of expired credentials reported the same outcome whether it cleared them or found nothing there, and a delete aimed at the wrong path reported that it had removed something.

The answer now says what actually happened: a token that was there and is now gone reads as a removal, and a path that held nothing reads as having removed nothing. Neither is an error — asking for something to be gone when it is already gone is a request that has been granted — and the platform has behaved this way since deleting something already deleted stopped being a failure. The token screen simply had not been passing the answer along.

This is the same promise the platform makes everywhere else a delete can land twice: a deletion that found nothing to remove still records that it removed nothing.