I don’t disagree with you, but I think the only fallacy here is that it’s an extremely zero-sum way to look at things.

Are we perhaps better off for many — maybe even all — of our status quo legal contracts not working like software programs? Sure.

Is there a class of legal contracts — either already in existence, or made possible by crypto — that’d make much more sense if ran like software (with different requirements/constraints than the error tolerance you described)? I don’t see why not, and why this would be mutually exclusive with the first premise.