Interesting read, but I feel like they should have also benchmarked using COPY with Postgres. This should be far faster than a bulk insert, and it’s more in line with what they are benchmarking.

The omission feels… odd.