You are seriously giving authors too much credit. The vast majority of the journal papers require enormous amount of production editing -- that's exactly why the EICs and societies sold those journals to publishers.

Things that I have seen : ( wife was a production editor, I briefly helped an EIC of a society journal that was sold to a publisher )

1. Graphs do not fit the page or space allotment - journal is uses two column format, author decided it did not apply to them. Graphs do not fit the color scheme - the author decided that the graph will be blue and the legend would be green. This is a black and white journal. Graph is a photograph of a graph, not a graph. Photo was taken on an iphone.

2. Sections went out of order. Author did not use the numbering system and numbered them by hand. Author cross referenced sections by hand, changed the ordering and did not update all of them, referring to non-existent sections.

3. Footnotes referred to non-existent papers.

4. Author included someone else's figures. Author did not obtain a permission to use them.

5. Accepted paper is written in a monospace font in WordPerfect for DOS. In 2014.

I’m not sure about other fields, but in computer science (or more specifically AI/ML) conference papers are completely free of any “professional” editors — burden is entirely on authors to handle their own editing. But I have noticed 0 disparity between conference and journal papers in terms of copy-editing quality. And my one experience of submitting to a journal matches other descriptions — the “professional” editors introduced spelling, grammatical, and even factual errors that weren’t there in the first place. (Most egregiously, they edited a grad student’s bio to award him a PhD that he didn’t have yet. WTF.)

Finally something useful they did -- helped the student :)

1 is mostly things that the publication team is asking authors to fix by email for modern journals. Most journals will have a template for these emails, or stick them at the back of the reviewer requests.

2 Is something that I have seen caught in peer review twice, and missed once. Again, it would usually go in a batch email to the authors of issues that need to be fixed before publication.

3 is definitely something that reviewers should be catching. I don't know of any journals (especially not open access) that check references for existence. Is the problem the use of footnotes? If so, I could see this legitimately requiring some manual labor / typesetting... But most journals forbid footnotes in their submission guidelines.

4 is quite rare in research articles, and again, would be handled by requesting the authors deal with it.

5. Would be grounds for paper rejection, as it breaks the submission guidelines for most journals. In addition, it seems like an exceptionally rare case.

Overall, I don't think one can justify an average of 10 hours per paper in manual labor (open access fees are always more than 1000 usd) based on these formatting issues...which issues are mostly handled by requesting fixes from the authors.

Then again, maybe your wife worked in the 1 percent of journals which actually print on paper, and carefully proofread for authors.

My background is mostly in medical engineering with journals up to IF10, though.

If this was the case then the journals would not be sold to publishers. They are sold to publishers because EICs and societies that start journals do not want to deal with actually making the manuscript publishable after a year or two.

P.S. Average STEM paper takes about 30 hours to "produce".

This was the case 30 years ago when publishing was expensive. Today this should be a button click in a fully automated system.

Which is not to say, societies don't benefit from the system -- they actually get a lot of $$ from it. The sad part is, they only get a small part from the much bigger $$$ going to publishers.

It is not. Publishing industry, especially academic journal publishing barely changed in last twenty years. At best there's now a workflow management systems that exist inside the publishers.

The bottom line is: if what technies claim publishing is could be was the case then the EIC/Societies would be publishing journals themselves.

Publishers are the AWS/GC/Azures of the world. They get to collect money from those that publish papers and need papers because they provide service that the paper writers can't seem to figure out how to do themselves ( it is not very surprising - quite a few of even the well known authors of well known papers even today insist on proofs being sent to them via fax, correct those with a pen and fax back corrections ).

> You are seriously giving authors too much credit.

Sure, how can we credit someone who did the research and prepared the paper in its publish-ready form :)

> The vast majority of the journal papers require enormous amount of production editing

Absolutely! Such as introducing new errors by people with no subject understanding. :)

> that's exactly why the EICs and societies sold those journals to publishers.

They are sold to them for the very simple reason: $$$ they get from it :)

> Graphs do not fit the page ...

If any of these are not acceptable, the journal is perfectly correct to request changes from the author. Which I as author would prefer over introducing damages by people with no idea what my papers are about.

There's no such things as "a journal" that "goes to authors for requests". It is a job of EIC, which EIC does for a journal that self-publishes. EICs hate doing it. That's why they sell their journals to publishers.

these do not require "enourmous" amount of editing. Still, if sience was published in open access HTML, none of these issues would exist. Graphs could be anything, layout is irrelevant, and the rest can easily be detected with an automated validator.

Scientists who claim to want to publish it as open access HTML do not actually want to do it.

Are you a scientist hating the publishers? I have a great solution for you. Publish it on your blog. Give the journal a limited reproduction right and make them put your blog url in the abstract.