True, it will (probably) be rare for my users to request a ‘second page’ as opposed to ‘give me a consistent sample of my results’ meaning of using first: without ever requesting a second page.
how do you know if you are at the end of a pagination result without doing an additional count every step of the way?
Maybe this is something only the database itself could do efficiently, since it could know if it reached the end of the tablet or ditched early (not saying they track that now but it could). Or, stream back results to the originating peer and client cancel the grpc stream when pagination requirements are met - or the stream terminates on the far side upon tablet completion.
A some level, I have to think that my example dataset of over 6M truly did not have to be completely iterated to get 12 results. After all, Only a Sith Deals in Absolutes.