LaTeX on its own does not magically solve all layout issues, especially if the exact placement of figures is pedagogically important or if the math equations are complicated. Usually you would end with 20ish compilation warnings (not errors) on a first draft. I think it is fine for a scientist to not bother with those, but I also think it is important for them to be fixed by someone in the final product. You can argue that I am snobish about typesetting, but that is different from saying that good typesetting does not involve an effort.

Also, I have never sent a PDF for publication. I always send source files. I do not know of journals that would accept a PDF (except for early peer review).