I've just had a problem with Compass connecting to an Atlas cluster. It appears to be related to some sort of DNS caching (either as macOS or inside Compass itself).

The steps:

1. Try to connect to Atlas via a SRV record, after ~20 seconds I get querySrv ETIMEOUT _mongod._tcp.<CLUSTER>.mongodb.net (see screenshot)
2. Tested via Mongo shell and SRV and can connect fine
3. Tried different Wifi networks, no change
4. Restarted Compass and used the same Favourited settings, and can connect now.

The Atlas cluster is an M0, and we checked the logs. Only interesting entries were that the cluster was hitting weekly bandwidth limits and was being throttled.

I also changed the password via the Atlas UI today, so it's possible that somehow that got Compass into an unstable state.