>
> I think an easy approach would be to COPY the CSV files into a separate database using psql's \copy command and then pg_dump that as separate insert statements with pg_dump —inserts.
>
This was my first thought too. However, as I understand, pg_dump --insert basically runs INSERT INTO ... sql for every row.
In other words, each row is un-prepared and executed individually. That is also not real life scenario.