Once a certain limit is reached, the AI stops working entirely

I don't know what caused it—I gave it a simple task to continue the previous one, but it just froze, so I had to type /new, and after that everything worked fine. I was using the Claude Fable 5 model; maybe they're having some issues with it, which is why this is happening—I don't know.

The report identifies no file, test, or entry point. Start by reproducing the freeze when continuing a task near the reported limit, using the Claude Fable 5 model if available, and compare it with starting a new session via /new. Done means the trigger and recovery behavior are isolated well enough to identify the affected component.

Written by the indexing model from the issue text.