On Postgres, I recently saw a case where there was a huge amount of selects being handled in a single transaction. This was really slow, so the solution was to break the selects into multiple transactions, which dramatically improved the query time.
Has anyone come across such a case? It seems counter intuitive and I was wondering if there was a technical reason as to why multiple transactions are preferable?
My only theory is that it's easier for the DB to maintain ACID properties with multiple transactions as opposed to a single gargantuan transaction.