Really fast batch superset count queries

We managed to come up with what we think is a pretty novel algorithm for computing vast numbers of superset count queries and would love some scrutiny to uncover if there's something better we could be doing.