To quote from the article: 'Electronic voting machines can be hacked, and those machines that do not include a paper ballot that can verify each voter's choice can be hacked undetectably. Voting rolls are also vulnerable; they are all computerized databases whose entries can be deleted or changed.'
It doesn't need visible chaos or detection on voting day.
A lot of the paper trails aren't necessarily even in machine-readable form. We are drastically unprepared for a large-scale recount (anything more than a county is going to take weeks+).
Also, the backup is effectively the checksum here too - unless you actually look for an error, how would you ever know if one occurred? We need x% of counties to be randomly recounted from the paper trail to validate the tabulations.
What measure of proof do you require?
Voter Action proved in a court of law that Kerry won New Mexico. No votes cast on spanish language were recorded. Hacking? Bad code? Bad set up? Either way, what's the difference? Regardless of intent, outcome is the same.
Voter Action chose NM because its electoral votes wouldn't the outcome (less controversy) and they had ample evidence.
This is not a unique story. In fact, its depressingly common.
Edit: Sort of a form of survivorship bias going on. The ones that "survive" (i.e. get found out) can't be used to draw conclusions about the others.