This is very interesting, because for me it's just the opposite. In particular the two column layout is just more readable and approachable for me. The PDF version also allows for a presentation just as the authors intended.
I guess it's good that they offer both now.
Hilariously, I would probably tolerate the HTML version a lot better if it had the font from the PDF (and FWIW, the answer for me is "no: I don't work with LaTeX at all... I just read a lot of papers").
Computer Modern was not designed for easy viewing on screens (think about the screens Knuth would have been using in 1977), it was designed for printing in books.
Not on arXiv (unless I'm much mistaken), which is a preprint server, not a conventional journal.
arXiv accepts various flavors of TeX, or PDFs not produced by TeX [0], and automatically produces PDFs and HTML where possible (e.g. if TeX is submitted). In the case of the example paper under discussion, the authors submitted TeX with PDF figures [1], and the PDF version of the paper was produced by arXiv. The formatting was mainly set by using REVTeX, which is a set of macros for LaTeX intended for American Physical Society journals.
Looks nice but seems strange to switch from two columns to one column after the first page? Although maybe they’re just trying to demonstrate its capabilities.
You typically send a .tar.gz of tex files (and, figures, .bbl, etc.) to the journal. And then you typically upload something very similar to the arxiv (I have an arxivify Makefile target for for my papers that handles some arxiv idiosyncrasies like requiring all figures to be in the same folder as the .tex file, and it also clears all the comments; sometimes you can find amusing things in source file comments for some papers).
Some fields may use Word files, but in most of physics you would get laughed at...
It is true that most journals will typically reformat your .tex in a different way than is displayed on the arXiv.
Not only is this wrong about physics/astronomy, I regularly use the arxiv version because the typography is better (e.g. in the published paper an equation is split with part of the equation being at the bottom of one column, and the top of the next, whereas the equation is on one line in the arxiv version).
