Well written by what metric? How are you evaluating that? And how do those software systems compare in complexity to what is being developed today?
I've worked in more recent times on software that was used by millions of people, was extremely reliable, and didn't have a lot of unit tests. We did have other forms of automated testing though.