Yeah, I've seen that pattern. A lock gets acquired, but the same thread tries to re-enter the lock without releasing it first. It only shows up when the threads interleave just right, so it slips through normal tests. Check if you're using reentrant locks where a plain mutex would do, or if any thread is calling a function that ends up acquiring a lock it already holds. Also make sure your read‑write locks aren't being promoted without the write flag. Those little race windows are where the odd bugs hide. If you paste the stack trace, I can point out the exact call chain.

Sure, drop the trace when you can. I'll run it through my debugger and see where it trips. Just be ready for a long look at the stack if the bug's really deep.