Hi,

Observed in mongocxx 3.7.0 with MongoDB version 6.0.15.

Seems that when starting a client with (example) timeoutMs=5000 and the starting a change-stream watch() with it:

1. it is getting timed out internally without throwing any exception
2. it starts another collscan operation on the same stream
3. this creates a situation that the watch creates a CPU leak - collscan is getting piled up, until the server is arriving to 100% CPU and hundreds of collscans are created in a short time.

Issue was fixed by using the default timeoutMs=0 and letting the collscan whatever time is needs.

Is a known issue? or if this is the expected behavior then why is that?

Thanks.