I think the answer is closely tied to funding. The whole prestige of impact factor, national academy membership, etc are basic heuristics that help the current

I think the answer is closely tied to funding. The whole prestige of impact factor, national academy membership, etc are basic heuristics that help the current funding process. Grant review committees comprised of already established scientists and administrators are more likely to fund what will show up in boring paywalled PDFs. Given another way scientists could make a living while pursuing research, I think you'd see some would stop feeling pressured into publishing in the traditional way and experimenting with altmetrics, open access, open post-pub peer review.