ShardLocal retries WriteConcernTimeout failures, which is at odds with both mongos and the drivers. We should reevaluate whether this is intentional and desired, as it isn't standard for clients to retry this error as its meant to gate the time waiting for replication. Retrying in effect doubles the timeout.

https://github.com/mongodb/mongo/blob/7a52188e306d69aacdb1175fd020aa2ed7e963ee/src/mongo/s/client/shard.cpp#L110-L112