You are right. Do we need Newtons and Einsteins in software development? Are there more or less objective (or at least consensual) indicators, that enough understanding has been built, and enough assumptions unearthed to start development? I mean, when one has a lot of experience, then it's not a problem. How to help juniors overcome naive realism?

As a side-note, non-techie stakeholders do not always understand the significance of assumptions, even if the latter are brought explicitly eg in negotiations. It does not help much to say after - "look, we told you so!" - as the client kind of is not supposed to understand matters deeply (that is why the client asked sw devs to solve the problem)...