And I say that as a long-time technologist.

Part of the reason that it’s the wrong answer is that
the wrong question is being asked — or perhaps another way
to put that is that the wrong
requirements are being articulated. Consider, for example:
what is the functional requirement for the time lag
between end-of-voting and result reporting?

I submit that the answer to that question ranges somewhere
from “days” to “weeks”, as in most countries there is a considerable
lag between when an election’s winner is determined and when
that winner assumes office. There is thus no functional
requirement for real-time reporting of results. (Yes, I’m aware
that the media would like this, but elections are not run
for the benefit of media. It is vitally important to get the
numbers right; it is of no importance at all to get them quickly.)

This line of reasoning (and others similar to it) lead me to
the conclusion that very simple voting systems (e.g., pencil
and paper) will meet all functional requirements. In addition,
such systems are well-understood — in part due to long
experience in the field — and numerous anti-fraud techniques
are known for them.

The answer isn’t to “fix” these hopelessly-broken
voting machines; the answer is to dispense with them entirely.

(Let me anticipate a possible counter-argument about
the tedium of having tens of millions of ballots repeatedly
hand-counted over a period of (say) a few weeks: don’t you think that democracy’s worth
that minor delay and trifling inconvenience?)