> Even in the modern era we still fuck constructions up, despite all we know.
That's precisely why we fuck it up, because we know tolerances and such much better, and we can do things with much finer granularity. So we aim to build right at that precise edge of "just barely above what's minimally required for it to work." And when small fuck ups happen, that edge gets lost, and suddenly it is all falling down.
In Roman times? No precise tools and techniques were available, things had far less granularity. So when they decided they wanted to build things "good enough", there was no precise edge. They would often ballpark for a construction project to be much stronger than minimally needed (by modern measurements), precisely because that "minimally needed" wasn't nearly as simple to achieve or even calculate.
Think about medieval brick castles. Some are still standing, while some houses built only about 100 years ago are pretty much disintegrating. Is it because we don't know how to build castles anymore? Not at all, it's just we try to build on the edge of the calculated expected lifespan of a house, balancing pricing at the same time. It was extremely difficult to do something like this with the knowledge available back then, so if they wanted something actually stable, they would often go for an overkill.