The last step of addShard in 8.0 and removeShard in 8.1+ is updating the "hasTwoOrMoreShards" cluster parameter if the number of shards goes from 1 to 2 and from 2 back to 1. If there is an interrupt before that step and somehow addShard/removeShard doesn't get retried (e.g. the router exhausted the number of retries which is 3), the "hasTwoOrMoreShards" cluster parameter would be left with an incorrect value. In 8.0, "hasTwoOrMoreShards" is used to determine if the directShardOperation check should be performed on a shard. If "hasTwoOrMoreShards" is false when the number of shards in the cluster is greater or equal to 2, direct commands against one of those shards would incorrectly bypass the directShardOperation check.