Thats why any person can act as an observer if they so wish and oversee the person counting your ballot. I can not possibly do the same with a turing machine.
So why can't you be an observer of the sys admin when they install the software. Witness some chain of command that the software has come from an authorized source, meets a digital check-sum and is installed properly. It can then be secured by another party with a two password "lock box" type approach kind of like we use with our crypto system at work that guards our CC processing. There are systems invented and implemented in the world already that can rid your concern over some rogue sys admin. Takes all of 1 minute of critical thinking.
You don't seem to understand, so heres a whitepaper [1] from Rop Gonggrijp on how to play chess on a Nedap voting computer. And the Nedap machines are special-built and programmed to ensure the authenticity of the votes, using anything from printers to read only memory. Not fucking Python running on a loosely secured Linux box.
Despite the halting problem being unsolvable, formal verification is a huge area in computer science. News flash: you can not formally verify some web app running on Python running on Linux, being fed with input from a gigabit link to the complete outside world.
So the idea of using Python for an electric voting machine is dead from the onset, a complete no-starter.
(And yes, of course computer languages are inherently more or less secure by themselves, through the simple proxy of allowing programmers to make (or prevent) severe errors. Think C and fixed-size buffers on the stack.)
