A while back we ran into an issue where we ran out of connections in MongoDB Atlas. It took quite a while to determine which database was being overrun with bad connections. It would be very nice if the API, which we use to pull metrics into another monitoring platform, allowed us to get certain metrics at the database level instead of the overall system level. The metrics I am suggesting are
Active Connections
Response Time
Failed Connections
Number of requests
Operation Execution Time

These would help us more quickly discover the source of the problem and trace it back to the application which is the initial cause