These compromises you brought up are a critical point and part of our most basic considerations. The approach of both Penpot’s Flex and Grid layout tries to bridge the gap between design and code through a shared language (wording and manipulation type, as you very well mentioned). With this in mind, I’m also well aware that this is a risky approach because, in some way, it could be said that it goes against the very reason why GUIs exist, which is (sorry for the extreme reductionism) hiding the computer complexity to match users’ mental models. In any case, both are features that are considered optional in a design workflow, you can do perfectly professional work at Penpot without them and under the usual conventions.
Our hypothesis is that in this specific case, bringing the two mental models closer together can be mutually beneficial, and we are already collecting evidence that shows that, at least with Flex Layout, the approach has been proven successful. However, Grid Layout has a higher level of experimentation and complexity and after various feedbacks like the one expressed above, we are already working on some tradeoffs to make it more understandable to people without previous CSS grid knowledge.
