Currently, adding a replicaset that already has a shardIdentity to a cluster results in a collision between shardIdentities, making the config server hanging and requiring a manual intervention. This situation arises when a shard is removed from one cluster (the operation does not clean up the shard metadata) and is then added to a another cluster. 
To address this, the AddShardCoordinator should verify during the kCheckShardPreconditions phase whether there exists a shardIdentity on the replicaset. If it does, the coordinator must fail.