i think we may have used an unpaged LR for certain endpoints and then counted the number of results, but only for use cases that we knew weren't likely to blow up