Running move/split/merge chunk operations as well as operations such as drop shard during defragmentation should restart the defragmentation, but should still end with successful defragmentation. We should create a test that runs full defragmentation on a large, fragmented collection with concurrent operations and ensure the end state is correct.