This realization is what has attracted me to the Erlang/Elixir world where fault tolerance > provable correctness.

The end user would rather have an application be frantically logging errors in the background, but be working for their usecase, than a program that is fixated with being correct and refusing to run in the presence of errors.

Obviously quality code still matters, but at least in the business realm "does it do what I expect?" is basically your success/fail state.