[wlonkly]

I believe it was I/O bound, on old, failing SSDs. Even after migrating object storage out to DigitalOcean's S3-equivalent, Postgres couldn't keep up on the terrible disks.