I will admit, when it came to brainstorming sources of crashes with threads, AI has helped me find sources I hadn't considered (as a systems guy, multi threading real experience is something I am sprinting through)
It looks like it wasn't too painful. I just kept a couple of array references too long. The LLM gave me some threaded crap, and I now need to make sure to let go, before another thread grabs the PC.
That's the issue with the new "memory is cheap" model. I cut my teeth with 256 BYTES (not KB or MB) of space for the program, the stack, and the scratch.
I suspect the best solution will be architectural, which promises to be a pain.