SERVER-65666 addressed this problem by making it so that the initial split policy will not choose draining shards when creating initial chunks. However, the create collection coordinator creates the initial chunks before actually committing them on the config server.

This means that there is still a case in which shardCollection will create chunks on a draining shard if removeShard is called after the initial chunks are created.

As part of this ticket, we should investigate whether this is additionally a problem with resharding. Resharding using the initial split policies to choose the new chunks for the collection, so it is possible that there is a similar problem here if the initial chunks are created and then a shard is removed before persisting those chunks.