In our test environment, we have our Node application deployed in our DMZ, resulting in there being a firewall between Node and Mongo. We've been seeing intermittent issues where Node floods the firewall with too many connections to Mongo (over 10,000).

We run a 3-server replicaset and have determined that all of the connections are always going to the same server in the replica set. That server, has been stuck in the RECOVERING state for weeks. So it seems there is some bug in the node driver that causes it to continually establish more and more connections to a server that stays in the RECOVERING state for too long