A developer debugging an autonomous sidewalk robot spent eight hours hunting a freeze that occurred 16 seconds after a client connected. The control loop, running at 100Hz, would simply stop without crashing. Standard fixes like switching to `std::thread::sleep` or swapping mutexes failed, leaving the system consistently deadlocking at iteration 1,615.

The breakthrough came with a heartbeat thread, which revealed the process was blocked, not slow. GDB showed an unexpected Rayon thread pool from the Rerun SDK, which the developer used for telemetry logging. The root cause was a known deadlock: calling into Rayon while holding a mutex, causing work-stealing threads to get stuck waiting for the lock.

The fix was deceptively simple: reduce the lock scope and avoid calling Rerun's `recorder.log()` while holding the mutex. This highlights a critical lesson in Rust concurrency: dependencies like Rerun can introduce their own threading models. The developer submitted a documentation PR to Rerun, hoping to warn others before they encounter the same bug.