Helpful on truly large query sets, or when you cant do as many splits. If your submit to a job scheduler, and your job gets killed for any reason, it would be nice to be able to reuse successful steps of previous runs.