An incredible amount of my debugging involves opening a CSV in Excel, saving it as a UTF-8 csv, and then just literally doing read_csv() |> write_csv() in R.

This cleans up all kinds of junk that seems hard to catch from various CSV exports that come from … out there.