Since the shard identity document is only written with w:majority (which doesn't even wait for the document to be applied on secondaries) it is possible that a mongoS contacts a secondary with a versioned request or some sharding specific request but that secondary does not yet know that it has been added to a sharded cluster.

This ticket is to throw an error in these cases and wait for the op time on that node to reach the latest seen topology time. Because the topology time is inclusive of the write of the shard identity on the replica set, this will imply waiting for sharding to be fully initialized. After this wait, the operation can be retried.