This is difficult in practice. For LaTeX, in theory the publisher would simply provide their style sheet (.cls) and maybe some style guidelines, and all the authors have to do is to adhere to that file and typesetting is done.

The reason this doesn't work in practice is that authors don't always play nicely, not because of bad intentions, but because they don't want to cooperate but because of the realities of life: they don't have the time to study style guidelines in detail, they use their own auxiliary LaTeX macro collection because that's what they're used to, or simply because of oversights. Also, typesetting often includes a whole lot of meticulous things, if you listed them all in a guide sheet, that would be a long list of stuff at a level that's too detailed for authors.

I'm not saying it's impossible for authors to fully follow a publisher's style guide but there's a reason publishers employ full time workers who do nothing else but correct submitted manuscripts. Like many other professions, it's a trained skill.

Nonsense. Formatting demands make things worse here, you could just ask authors to submit unformatted content (e.g. Markdown or RMarkdown, or utterly minimal LaTeX file, with references and a bibliography file) and then trivially move that content into whatever format is required. There are in fact journals that do this too (i.e. don't have formatting requirements).

As a submitter applying to multiple journals with arbitrary formatting requirements, you are often forced to meet arbitrary and irrelevant (visual) style requirements even before you are likely to be published, so of course you keep a base unformatted copy that you modify as needed to satisfy whatever bullshit policies each random journal demands. This wastes everyone's time.

The reason submitters don't "play nicely" is because the publishers' demands ("style guides") are demented here: they should just be asking for unformatted content (besides figures), certainly for submissions, and even for accepted publications: they should actually be doing the work of formatting and typesetting. But instead they force most of this on the submitters, to save costs by extorting the desperation of academics.