For scale: extrapolating (badly) to the whole US registered voting population (~175 million) = ~5,000 "likely noncitizens" who voted at least once since 1980.
I'm curious what the numbers are after they've confirmed the accuracy of the results (not just "likely noncitizens).