https://github.com/mongodb/mongo/blob/1f6d847e7f9ac00b29e43214769f29a15017365b/src/mongo/db/s/resharding/resharding_data_replication.cpp#L372

SERVER-113039 implemented a workaround for SERVER-103945 by creating new child factories to limit the lifetime of attached cancel source and it's listeners to the child factories. We should consider creating a new hierarchy so the cancel source listeners can be cleaned up when the various data component finishes (the cloner is probably the most relevant since it runs the longest and also in the earlier stages).