Having helped with an e-election system myself, I saw first-hand how it caught "bugs" in the process. For example some district entering wrong information (lots of cross-referencing checks tripped an "alarm").
In the U.S sure, I can see why people would be concerned. No offence, but the systems in place for social and other citizen-related info ain't excellent. If there's an entire market for false identities it's saying something. Taxes not being done automatically. Social security number being the one way to identify. Online banking being a pain in the ass.
Scalability is an issue. If a system is open to billions, there is more incentive to work on "theoretical" exploits. But let's not pretend paper-voting is a better alternative. E-voting doesn't solve the corruption problem, but it makes it easier to find.
Tom Scott and some of the pen-testers that shit on on the concept have good points. However they all are based on the idea that staffers manually counting works better. Tom Scott's New Hampshire example is anecdotal – that system was an insult to the word naïve. The pen-testers taking a dump on Estonian system were picking on stuff like WiFi passwords being visible and seeing over the shoulder an admin's terminal. The systems responsible for the counting weren't connected to the WiFi and it was there for guests. Lot of good staring at a terminal did with no access to the actual machines holding the program.
I don't see e-voting becoming a thing due to all the FUD spreading. But I hope it will be reviewed as a means of double-checking. Perhaps some studious people might actually go out and study the actually proposed architecture. It's really never as simple as one program doing the counting with a flavor of auditing. At least when done right(ish).