When a replica set connection fails the error message can be fairly opaque, requiring lucky guesswork or heroic efforts to fix. A common problem we have been running into is connection timeouts from Heroku as shown here:

https://github.com/jcottr/temps-mort

It would be nice to have better tools for diagnosing the issue in these situations. Either better error messages (in this case, it would have been helpful to know that we were getting timeouts) or maybe a diagnostic mode you could put the driver in where it would log diagnostic messages including the individual errors it gets on connection attempts.