I find it hard to believe that a modern identity infrastructure (which we don't have, admittedly) combined with basic cryptography can't get us where we need to be.
We might not be able to cryptographically prove anonymity, but all it takes is trusting the government to anonymize the data correctly and make it secure in transport.
Large banks don't always succeed at that. The government makes huge mistakes all the time. What makes you think this is possible?
The issue is that perfect security is impossible either with physical or digital voting. All we can do is minimize the consequences of an inevitable mistake or breach.
The point is, computers provide a single point of failure. Bribing enough of a country's electoral officials to shift a vote without being detected is immensely more difficult in comparison to finding a zero-day in some voting software.
The only guarantee that open source provides, is that we'll probably find the bug eventually. A malicious state actor that wants to influence an election has absolutely no incentive to let people know that they've found a vulnerability with any sense of alacrity.
Wait, what? All the system requires is trusting the incumbent government with the security of our vote?!
And the fourth Reich lasted forever.
Have you read "Reflections on Trusting Trust"?
You have to make some higher level assumptions in practice or you won't get anywhere.