This removes a lot of overhead when there are node groups that are reused a lot and only need to be processed once.