At the same time you have parts of a country that actively suppress votes. For example, in many deep south counties that are turning blue, the long standing majority would make up the election commission staff and other election related positions. Organized fraud is hard, but if enough people agree that fraud is neccessary, fraud will happen at a large scale.
I am not saying digital election is better, but I would like to see a formal securirty/risk assesment of both options against known voter hostile counties.
With mail-in ballots for example, what if small groups of mail persons(1-5) manipulate absentee ballott mail when sent to and collected from voters, how does that risk compare to say insecure network connections and poor authentication with digital voting systems. A formal threar model for both approaches and risk assesment that is. Both methods are not secure, as a techie, the vulnerabilities of digital systems is glaring and obvious but they mean little in terms of risk wihout applying a threat model.
You said paper ballots make mass fraud hard, that is because of the threat model you had in mind. The chaos of the 2000 presidential election shows a slightly diffrent threat model where organizing large number of people was not needed. Bad guys will always find a way, what I like to do is to take an empirical approach at making thier mischief harder to pull off.