Maybe full hand count is not feasible. Supposedly the UK figured out how to hand-count their BREXIT referendum, so it might be possible. I'd rather throw a few B$ at the problem than deal with what we've got right now.

I guess the most important thing is to keep the hardware and process as simple as possible, to eliminate the complexity and black boxes that can be compromised, and to give the best shot at being able to verify it. Having standard hardware that has well defined functions and never gets updated, would also help. I'm thinking standardized on a fairly simple, bulletproof scanner unit, with open published designs, firmware, etc.

Openly publishing high quality ballot scans (showing handwriting artifacts, etc) would itself go a long way to securing the process, since anyone online can vet it on their own time. If a problem can be observed there then you delve into an audit.

I don't think it's strictly true that quantum can crack any encryption. (One time pad for example). It can definitely crack many forms of encryption though. But I think my position is that faking "real life" so well that a good investigator can't catch it, is very hard.