The shard ShardRegistry should mark hosts which failed due to OperationTimeout as faulty in the ReplicaSetMonitor so that they are not being contacted again on a second attempt.
If the host timed out because of some transient error, it will eventually be put on the online list again and will start being contacted.
- related to
-
SERVER-22620 Improve mongos handling of a very stale secondary config server
- Closed