@spectral sleep until timeout is up is dangerous as it greatly increases the concurrency of the system when the lock is hit. better to just fail fast and re-enqueue