OTF sampling has an optimization where we only retrieve fields in the sample that are relevant for the query. This code calculates which fields those are. We should re-organize the code so that we don't do this calculation unless we know we will do OTF sampling since this information isn't actually used on the persistent samples path.