Publishing scientific papers should not be profitable or even financially sustainable on its own.

The activity we need to fund is the research itself. If we don’t allow profiteering from publishers, let’s not pretend would ever see the papers or that the research wouldn’t happen.

Uh, they don't fund the research, they don't pay the reviewers, and they charge money to publish. They provide exact formatting rules and documents for all submissions, and reject if improperly formatted.

I'm not sure where you get this idea that they are doing any significant typesetting, or that their administrative costs are anything remotely reasonable - even the journal editors are also largely unpaid (because being the editor of a journal can be helpful when trying to get a job in academia).

I think it depends on the field. In Computer Science, all of the points you make are hilariously wrong. They are true in older fields (though those are slowly catching up with technology).

If anything, CS journals are more old-fashioned than those in natural sciences.

My least favorite part of publishing a CS journal paper is checking the proofs (in typesetting sense). I have already typeset the paper using LaTeX with the journal's template, but the journal almost always wants to typeset it again using another system. That apparently involves a lot of manual work by their staff. In the process, they introduce many errors into the paper, some of which are random and others which are consistently wrong. I have to compare my PDF to the publisher's version visually to spot the errors. And there is usually only 48 hours allocated for that part, despite the publication process taking 1-2 years.

Covered by downloads? Definitely not, they're more than covered by the contracts that they have with academic libraries to sell back the published papers to the institutions

Of course most of the money comes from subscription fees, and of course both the publisher and university monitor the number of paper downloads. While the actual numbers are not public, it's reasonable to assume that the subscription fee divided by the number of paper downloads covered by it is within an order of magnitude of the price of an individual paper.

My point was that it's covered irrespective of the number of times a paper gets downloaded or read. The papers could get read 0 times but they're still making bank as long as they have those library contracts

And my point was that the average number of times a research paper gets read is low, regardless of whether it's behind a paywall or freely accessible. Given the costs of commercial publishing, the average cost of reading a paper must be high, no matter who pays for it and how.

It's far too high, the publishers shouldn't be able to capitalize on publicly funded research by gate keeping it