A Cute Bug in HyperQueue: SIGTERM and the Ten-Second Mystery

A curious bug emerged in HyperQueue, a Rust-based distributed task scheduler. Tasks, particularly those sleeping for more than 10 seconds, would mysteriously terminate. Debugging revealed a seemingly innocuous change: offloading process spawning to `tokio::task::spawn_blocking`. This, combined with `PR_SET_PDEATHSIG` (which sends SIGTERM upon parent process death), caused the issue. The worker thread spawned by `spawn_blocking` was being reaped by Tokio after inactivity, triggering the SIGTERM signal. The bug was fixed by reverting the optimization, highlighting the subtle interactions between concurrency, system calls, and thread management.