When server selection fails, we can improve diagnostics in two ways:

1. Check if client or cluster have been closed/disconnected. This will catch situations when the application closes the client and then tries to use it.
2. Check if monitor thread is alive. This will catch application not reconnecting the client after forks.