I have just submitted an article to an Elsevier journal. They have some non-functional web interface so that you can edit the article yourself. My biggest problem was fixing the tables, because they looked really bad after the automated copy editing. There was some text on the interface which said something like "This will not be the final version of the tables", or at least that's how I understood it. Well, I clicked submit and without any real editing from a human on their side the article was published. Now my article, which I have put so much effort in, has messed up tables.

Same with me. We had submitted good quality graphs according to the requirements. When it got published the graphs got compressed and look worse now. Funny thing is: one of the graph was not high quality(done in Word), that one looks much better.

Who cares, everyone interested will just read your pre-print. :)

Right. The reviewers who did the peer review were unpaid; readers will read the pre-print. What value does the publisher add?

In my experience they mostly complain about the format of references not exactly matching their special standard.

Right, and then they ignore the .tex sources, including formatting, and reset the whole thing in some weird proprietary system anyway, making half of the compliance to formatting that I did completely pointless. (And in the resetting, they introduce a few new errors into the references for good measure.)

Yeah, the whole problem is that traditional publishing is designed around creating the best experience for publishers, not for authors or audiences.

The reference formatting thing is just one of the ways they make it obnoxious for authors (it would be better for authors if you could just send them a list of DOIs and their computers should make it look the way journal style dictates — ELife does this). Idiosyncratic rules about the naming of sections or formatting of methods or supplementary material make transferring article between journals (even at the same publisher) unnecessarily tedious.

From the audience perspective, nobody wants to vault over the paywall to click on the link to click on the link to get the pdf that displays in a pane of the browser window. Nobody wants an enhanced pdf, whatever that is. I don’t want to see a pop up with the articles you think I should read next because they happen to share a single word in the title. I just want to click a link in pubmed or google scholar and go direct to the pdf. A few months back someone posted an enhanced google scholar that just linked directly to the PDFs from sci-hub. The user experience was so good that it really highlighted how obnoxiously bad publisher sites are.

Off-topic nit: the word "painstaking" is a concatenation of "pains-taking", i.e. "taking pains". It's not usually given a dash, but if you put one in, it belongs after the "s".

On-topic: And that's just the sort of copy-editing that would end up being done by (unpaid) reviewers and not the paid copy-editors.

(Thank you, by the way.)