For ETL out of Postgres, it is very hard to beat psql. Something as simple as this will happily saturate all your available network, CPU, and disk write. Wrapping it in Python helps you batch it out cleanly.

    psql -c "..." | pigz -c > file.tsv.gz