Apart from everything else, if you're not trying to
ensure that a candidate wins, regardless of their popularity, and are instead trying to skew the odds toward a candidate who already has a reasonable chance of winning, election fraud becomes easier and harder to detect.
As others have pointed out, these "judges" were actually minor, elected election officials, close to being volunteers. They were doing this to make small additions at a local scale. There was no need to add fake voter registrations, to modify vote counting, or to add votes not connected to legitimately registered voters: they just added ballots and records for registered voters they knew weren't going to show up. At a local enough scale, you might simply know, personally, of voters who are out of town, for example.
This doesn't require any major conspiracy at multiple levels. Depending on the organization of the election, it might be possible for a single poll worker to do it on their own. It would be very hard to defend against at higher levels. Most voter ID ideas wouldn't help (short of digital IDs and cryptographic signatures), because it's being done by the people who would be checking the IDs. Having multiple, adversarial officials keeping records of each person coming in could help, but now you've multiplied the number of people you need at each precinct. Contacting people listed as having voted could help, but they could well have been chosen specifically because they would be unlikely to notice or respond. Checking counts and registrations wouldn't help, because the counts and registrations would be valid. Voting technology mostly doesn't matter, and in fact, the method is likely easier with paper ballots.
It is limited in how much of an effect it can have, of course, but in tight races, or down-ballot races where few people actually fill out those races on their ballot, that might be all you need, or you might be interested in just statistically helping your party by making larger numbers of your party's candidates win, rather than helping one particular candidate.