Currently it is not possible to invoke a removeShard command on a specific machine and add it again at a later time without removing the datafiles due to the following error :

"errmsg" : "can't add shard localhost:27022 because a local database 'test' exists in another shard0000:localhost:27021"

I understand mongo cannot simply remove the database from the removed shard for a variety of reasons but I think the validation step to determine if the shard to be added is "clean" can be improved. I would assume iterating over all partitioned = true database on the new shard and for each iterating on all sharded collections to check if they're empty should be as safe a check and would allow subsequent removeShard->addShard without further intervention.