the current problem with journals is that there is no money in it for authors. journals oly exist because of historical reasons, and older folks still value them.
Arxiv exists as a semi journal, which is some what cc4 (or some other cc of your choice) and that is great, but still one source.
You can just host your research papers as websites, as in just a web article, and use some vcs like github, codeberg, or self hosted forego system. That is arguably the best case.
I have a paper which is on arxiv, and my supervisor has been "polishing" it for a journal, but to me that is a useless process, because i almost never care about things like journal impact factor or h index. to me, the only thing valide is steps for reproducibility, that is, give me a recipe, and if i can recreate, then you did a great job. This could mean, for example, releasing all your raw unprocessed data.
how to handle reputation for who can review, but I think there are ways to do that and that’s beyond the scope of this post as I imagine it could get pretty complicated and would require feedback from people actually in the industry. The reviewers can submit comments and reviews back to the author via federation, but this time the process can be open instead of behind closed doors.
one of the reasons reviewers are effective is that the remain anonymous, that is why they can shit talk a lot. You would not have the slander, if you make the identity real.
I think we should not have reputation or verification, as i stated above, if you post on your own website, and not have gatekeeping. Yes a lot of the work may not meet "some standards". but even with current system, a lot of work is published which is substandard. if we can release work in open, and colaborate as we do for open source software, thart would be the ideal thing for me. Each issue could be a literal git issue, each correction can be a pull request, and so on. Fully transparent, and somewhat resistant to whole network failing. (assuming you have local copies, you can just spin another instance, and your paper still stays onloine)