We've seen the same process in the pharmaceutical and newspaper industries, and likely others.
While I agree with pretty much everything else you wrote, I have to take some exception to this. Yes it is trivial to put something online. But library sciences exist for a reason, so if you want a nice collection of things online with some sensible structure it becomes less trivial. Worse yet if you want some way of ranking in importance the things in your collection, it requires a lot of manual intervention.
I think your solution is ultimately correct, but the peer review process is harder to solve than the technical one, and there is still a cost of some sort to implement and maintain the technical one.
Not publishers.
Libraries.
Publishers do not currently do this. Their websites are a confusing jumble with difficult navigation, poor search features, no useful discovery features, ...
The “nice collection” with “sensible structure” is provided by external citation indexes like Google Scholar.
All academic publishers contribute to the process is an early-2000s web interface.
Specifically, what structure do paid journals have that Arxiv lacks? How much would it cost to provide that structure outside of the current system, bearing in mind that editors and peer reviewers aren't paid a dime by the publishers?
The peer reviewers may not be paid but the prestigious academic journals (Nature, Cardiology, etc) do have paid editorial staff. E.g. the initial rejections of papers and rounds of back-&-forth copy-editing with the author are handled by the paid editors and not the unpaid peer reviewers.
That's not to say there are also journals that are run by 100% unpaid volunteers but I'm unaware of any particular Elsevier journals with that labor structure.
(However, this is not an endorsement of Elsevier's high subscription fees.)
Science and technology progress on correctness, accuracy, and efficacy, not popularity.
It turns out that gatekeeping, cabal-fostering, orthodoxy-preserving, and kingmaking are lucrative practices. But truth cares not for prestige. As the motto of the Royal Socity states: In nullis verba. By the word of no one. Evidence, not authority.
(The situation may be different in some of the arts and letters. Which speaks to another can of worms.)
To try to bring an analogy, D. Knuths recent comment on the recent sensitivity proof has brought the discussion of the topic on a different level because it has basically been stamped with the approval of a highly prestigious scientist.
So although the sensitivity proof is a mathematical proof, the recent buzz around it has rocketed it to another level. This brings about real rewards in the form of promotions and invitations to exclusive social circles, all of which are desirable to humans.
To be slightly cynical, the majority of research output is not worth your time to read, so mechanisms that presort (E.g. by author list, keywords, and journal name) help to increase the chance you read a paper worth your time.
Is prestige very much baked into the practice, profession, career, and culture of academia? Yes. Has it largely always been thus? To the origins of the modern university, if not before -- arguably.
But that's not a necessary role, certainly not for publishing to play, and though giving a thought to side effects and consequences, disrupting the publication cycle might solve a few more problems then it creates.
Or not.
The deeper problem is that demonstrations of correctness don't adhere to publishing, funding, teaching, or even career cycles. Things can be thought wrong for a very long time and proved right, or vice versa. And again, all the prestige in the world won't change the truth value.
Though it will change what truths, or myths, are socialised at any given point in time.
However, I do think it is getting better, and a growing number of scientists (particularly the younger generations) are pushing back.
The internet should have changed this model, but the publishers had amassed a secondary form of value - prestige. Scientists don't only want to share their research with their peers, but share that research in a prestigious publication. Getting a paper published in Nature or the BMJ has a meaningful impact on the author's career.
The internet makes it trivial to distribute documents for free, but nobody thus far has found an adequate substitute for the prestige of a highly respected journal. Universities keep paying to subscribe to prestigious journals because they publish some of the most interesting and significant research; academics keep sending their best work to the most prestigious journals because it's valuable to their career prospects.
This problem can probably only be solved through coordinated action by academics.