This has become significantly more complicated with successive iterations; see an earlier version in the history (like 2633d5b) for a simpler example.
Note the geographical caveats in the NYT dataset; the most annoying of which are the fact that the 5 NYC boroughs are all lumped together and Joplin and Kansas City, MO report thier numbers separately of counties they span. Further, note that US territories and the District of Columbia are listed as "states."
This dataset is joined with the 2020 US Census numbers as mucch as possible to facilitate rates per population, but special care is needed for some caveats. In particular, the city of KCMO is split between portions of 4 different counties (Cass, Clay, Jackson and Platte) and does not wholly encompass any one of them. Since KCMO is reported separately, we pull its population estimate from its metro region association's estimates, and then remove population as appropriate from each of those four counties.
Some states report cases from "Unknown" counties (with 0 population).
This was setup to be hosted on free-tier Heroku dyno. In the free tier, the app gets killed after 30 minutes of inactivity and we're given terribly anemic access to the CPU. To reduce startup costs, a special heroku buildpack that uses PackageCompiler.jl is under development.