Other sites permit one to append ".json" for a raw JSON feed.
There are sites already which provide some form of this, notably sites using MediaWiki (which notably powers Wikipedia and Wikisource) offer multiple download options, though I don't think it's quite as flexible as I'd prefer.
Client support would be the other factor, with current Web browsers not being particularly multi-format. Though to be fair, Firefox, Safari, and Chrome all now natively support PDF documents.
I suspect that a render-to-ePub would be a much better alternative to standard HTML for mobile devices. ePub is inherently fluid, and tends to offer a subset of full HTML capabilities. (ePub is of course an archive of HTML, stylesheet, and other elements, though usually it's more strongly textually oriented than the modern Web is.)
For sufficiently large tablets, PDFs are actually a quite good document format.
I've also found it ... interesting ... to note that a single Web page often has the memory footprint of an entire book (a few MB for just plain text, though that can balloon upwards with complex typesetting and graphics), and that even a browser optimised for e-ink (Einkbro) has roughly 10x the power consumption of the stock ebook reader (Neoreader) on my Onyx BOOX tablet (labeled as an e-book reader, but in fact an e-ink based Android tablet).
I can read books for days on a single battery charge. I can browse the Web for ... a few hours.