Though I understand the impulse not to make predictions
, I think it's obvious that eventually typesetting quality will approach parity with paper.
Ebooks currently make up a tiny slice of the market. In a few years, ebooks will make up a much larger slice; eventually they are likely to be the majority of sales. At that point, the ebook will be the primary emphasis, and paper an afterthought. It might even be possible that booksellers will develop some sort of format that will be possible to use to derive both ebook and paper / POD printed versions, thus reducing cost and time to deliver.
There may still be some issues, since you're dealing with a very different types of output. To get a paper book right, every aspect needs to be highly specifically determined. With ebooks, you're dealing with numerous dynamic elements to the output -- e.g. different sized pages, alterations in font sizes, users who will change the page orientation, etc. Other aspects like ligatures are utterly superfluous for digital output, and IMO it's fine to drop them. So I expect some of the fussier elements might fall by the wayside.
And not every commercial ebook today is a typographic disaster (though too many commercial ebooks have errors). I've purchased some ebooks that had rather complicated and well-executed formatting, and essentially none of the common errors that I presume come from automated conversions.