Agreed that if you want to keep the relational part, using the DB to store pointers is probably the easiest, though it might require rethinking search. I appre

Agreed that if you want to keep the relational part, using the DB to store pointers is probably the easiest, though it might require rethinking search. I appreciate you at least considering the various options. If you do nothing else, tuning Postgres parameters, optimizing column ordering, and removing duplicated indices will be a great step forward that is completely backwards-compatible.