> the typesetting done by a publisher is in the majority of the cases on a very fine-grained level. A lot of is is simply enforcing the rules that were missed by the authors (with the goal of fairness, comparability, and conformity) and small perfectionist's edits that you might not even notice at a casual glance but that typesetters are trained to spot.

As I said, if this is the case, the vast majority of typesetting and formatting has clearly been outsourced to submitters, and this means the amount of actual typesetting/formatting done by journals can only be minimal compared to in other domains.

EDIT:

> On top of that, authors often want to present their research in a specific way, and often have strong opinions about e.g. how their formulas are typeset, what aligns with what else, etc. and typically spend quite a bit of time tweaking their documents to look the way they want it. That is, the authors already have an interest in using something more powerful than Markdown.

Yes, generally, I don't want to present my formulas and figures in the shitty and limited ways the journal demands, but which would be trivial to present on a website (which is the only way 99.9% of people access articles now anyway). So journal requirements here are usually harmful and generally 20+ years outdated.