I might be biased on "majority of researchers read papers on computer". As to "high-quality typesetting", I think that depends on your standard of "high-quality". If the goal is "easier to read", then it doesn't have to be traditional typesetting. LaTex is not the only way to make authentic typesetting (and don't get me wrong; I use LaTex a lot myself). You can do that in HTML/CSS too. Actually HTML is more flexible on this because you can provide different views (e.g. Sans-Serif vs Serif fonts) for your reader for on-display reading and printed-out reading.
You can compile LaTeX to two different output formats if you want to. Sure, you'd have 2 different output files, but e.g. for slides, thats done very often.
Comments
I might be biased on "majority of researchers read papers on computer". As to "high-quality typesetting", I think that depends on your standard of "high-quality". If the goal is "easier to read", then it doesn't have to be traditional typesetting. LaTex is not the only way to make authentic typesetting (and don't get me wrong; I use LaTex a lot myself). You can do that in HTML/CSS too. Actually HTML is more flexible on this because you can provide different views (e.g. Sans-Serif vs Serif fonts) for your reader for on-display reading and printed-out reading.
EDIT: typo
If there were a browser plugin that applied Knuth's linebreaking algorithm for printouts, this would be more appealing.
You can compile LaTeX to two different output formats if you want to. Sure, you'd have 2 different output files, but e.g. for slides, thats done very often.