In SingleServerDiscoveryMonitor::_scheduleNextHello we will schedule a callback on our executor to send a hello. This is done via calling SingleServerDiscoveryMonitor::_doRemoteCommand(). _doRemoteCommand() itself schedules another callback. Let's call the outer layer callback A and the inner layer callback B.

If callback B's handle returns a bad status, _doRemoteCommand() will throw an exception from that status. This will then bubble back up to callback A.

This is a violation of the contract for TaskExecutor::scheduleWorkAt, which explicitly states that callbacks given to the executor must not throw exceptions.

This contract violation has caused a crash via the linked BF, and must be fixed.