A useful use of AI might be to simulate thousands/millions of user sessions (via generating mouse and keyboard inputs), with instrumentation/logging switched on. Run them in various shapes of VMs until you hit the problem. Fuzz testing, basically.
Which might hallucinate some more problems which doesn't exist in the first place?
Or it might find it. You don't know until you try, but for a bug of this severity it's worth a shot!