This is to fix the possible StaleConfig error a user encounters when upgrading a rs to a config shard. 

The idea is that after the CRUD operation initially fails due to StaleConfig, the sharding metadata will be recovered, and so when the driver retries it, it should succeed.