The failure lies in forgetting that computers are not deterministic machines, they are probabilistically deterministic machines instead. They are physical cons

The failure lies in forgetting that computers are not deterministic machines, they are probabilistically deterministic machines instead. They are physical constructs, not mathematical ones, and as such, verification is inescapable if you want some level of assurance. You need tolerances and experimentation after construction, even if you proved your thing correct. In the words of Knuth: "Beware of bugs in the above code; I have only proved it correct, not tried it."