In theory, yes. In practice, this only works when the document is using a restricted set of packages. I tried.
I'm fighting not so much with LaTeX, rather with ArXiv. They have the sources of LaTeX submissions, but do not provide other formats than PDF by default. Judging from my own experience in converting LaTeX to HTML, epub, etc., I assume these conversions simply create too much headache.
> reading a 25 page manuscript is not very efficient on a screen, at least in my discipline
The vast majority of papers I "read" I never read back-to-back, rather than Abstract - Discussion - Results - Introduction - Methods, bailing out at any of those points when I have the information I was looking for. I couldn't possibly keep up with the literature when reading every interesting paper completely.
Those that are really relevant to my research I will read in detail, and at that point a PDF is useful, I agree. But before that, I prefer reflown text & figures.
There was some innovation in that regard, e.g. eLife's Lens, that was specifically designed to support reading on a (large) screen. IEEE also has a good system that works on phones (well, almost; so UI elements clutter the text).
It's not about LaTeX per se. You can typeset EPUBs with it too, if you want. So reflow ability is not a shortcoming of LaTeX.
Publishers send you a Word or LaTeX template, and require your manuscript as a PDF, conforming to that template. If it's LaTeX, generally there's a switch to render the text in "Review mode" with line numbers and similar additional details, and omitting author data if the judgement will be done blindly.
So, the tool itself, regardless of its brand is the proverbial wrong tree to fight with. Also, it's a reality that reading a 25 page manuscript is not very efficient on a screen, at least in my discipline. I have a dedicated "Paper printer" at home for printing such manuscripts and taking a stab at it with pens and highlighters.