OK Datapedes that are familiar with the NYTimes JSON data! I don't know if this has been pointed out, but I noticed in the NYTimes source that there are data files for swing states, and these are very different than the general election total files. They contain ACTUAL NUMBERS of total votes for each candidate by precinct...not just rounded-off percentages! They apparently don't include every county/precinct, but much of what is there includes the state, county, precinct, vote type (absentee, in person, etc.)!!! The state totals seem to be well below the full state totals, so I'm assuming that a good bit of information is missing, but there may be some good data here that someone could find anomalies in. Here's an example of some tinkering I did in Power BI (EDIT: That middle image is Kent County, MI, Not PA), but there's so much info there that I don't really know what I'm looking for. The relevant ones seem to be for MI, GA, and PA. I've come up with a trove of the ones I could find going back to election night.

• Here is a link to a combined single .csv data set that includes the filename and a datetime column you can use to sort them (I left out the third party candidates because they vary between states, but you can calculate the third party votes by votes - bidenj - trumpd) (Another note: If you import this file, the second column is the time parsed from the filename. Somehow, the colons became underscores, so you may have to replace "_" with ":" in the second column to convert to zulu time in that column)
• Here is a link to the raw JSON files.