I might be mistaken but Sydney Opera House is brought here (as it often is) in context of project management as a heavily underestimated project (took 15 years to build while planned for 4, cost ~15x of original estimate) - not in the context how detailed and greatly architected it is.
There are deep analysis from the project management perspective but one point was that initial estimates were based on historical assumptions while some elements were novel and couldn’t be estimated with those techniques.
With software it’s similar. We have some general feel how things should work but often there’s “this” novel part that no one did before and it’s wishful thinking, leading to budget and timeline spillover.
If you ever had to guess-estimate something you never saw and it turned out to be overly optimistic - that’s similar case to Sydney Opera House (which I recommend to read about as there are many analyses of the case from various project-related points)