The nodes structure is initialized incorrectly in ReplicaSetMonitor that made them initialize to ok = false instead of ok = true. And because of SERVER-5405, the node can forever be marked as not ok. The side effect is that every time a connection needs to get the master, it will call isMaster & replsetGetStatus (in the normal case, will never get called because node is ok).

This also means that slaveOk will never work on a sharded environment with auth because the secondaries will be marked as bad nodes from the start and will never be able to recover from it (Without SERVER-5405's fix).