> That seems to be the crux - scientific papers will have to fall into two camps "blogs or basically deciphering the lab notes of everyone in your field" and "look this is a real effect and worthy of your attention"

This ignores reputation and expertise. If you’re studying the sociology of Hollywood there’s Gabriel Rossman and maybe six other people who do quantitative work. They all know each other and read each other’s work. If one of this small group has a grad student who does quantitative sociology work on rich datasets from IMDB or somewhere else they’ll be introduced at some point while doing their doctorate. He also partakes of other networks within sociology, Economic sociologists and Princeton graduates. All of these networks are relatively small and reputation and gossip travel quickly. People hear that others are good or not getting tenure, etc.

Sociology is a normal social science. Papers go from an idea to a sketch of an idea or maybe a poster to a conference paper or graduate seminar discussion to working paper before finally being published. The peer review step is the last one for the diffusion of knowledge about what’s good. If people stopped publishing in legacy journals tomorrow peer review would still happen, it’d just be post publication peer review of working papers, the system Einstein published all but three of his papers under.

You can get the second without personally wading through the lab notes of everyone in your field because if you’re trying to get attention you need to present your results and advertise them so people will care, unless they already care because you have a reputation.

For PNAS, I've heard, getting sponsored by a NAS member greatly increases chances for publication. So it seems an odd example to cite for "value".

Its more that NAS members can publish some articles basically for free, so co-authoring with them is a good way to get in easier.

That being said, NAS members I know do not want to erode their reputation by publishing junk.