> So, if the "stack" is a, say, finite resource, then the programmer should know in the code how much of that resource is being used and act accordingly.

And this is true, but IIRC statically determining stack bounds for arbitrary programs is not an easy problem to solve, especially if you call into opaque third-party libraries.

> For a preprocessor for C++, I IIRC at one point the definition of C++ was in terms of a preprocessor

I wouldn't know about defining C++ in terms of transformations to C, and searching for that is more difficult. I would guess that the abandonment of the preprocessor approach to compilation would also have meant the abandonment of defining C++ in terms of C, especially once C++ really started picking up features.

> The issue is that at least at one time it seemed difficult to be precise about C++ semantics, that is, what the code would do and how it would do it. Maybe now C++ is beautifully documented.

C++ has had a formal specification since 1998, which might count as documentation for you.