I'm trying to figure out why a full CSV export from https://covid-19.datasettes.com/covid/ny_times_us_counties runs unbearably slowly.

It's because the streaming endpoint works by scrolling through every page, and it turns out every page calculates facets and suggested facets!