Platforms like Authorea and Overleaf combined with preprint servers can help society journals compete with glamour journals without going broke. Essentially the society journal is destined to become an "overlay" journal where the editor suggests what preprints to look at while they're under review. This should, in principle, encourage actual review. Open review offers another type of incentive. Of course the actual society journal editors can be dinosaurs (and I say this after speaking directly with some of them), but eventually most of them get with the program.
I don't know what to do about the glamour journal fetish. Many of my papers in non-glamour journals have dozens of citations, which I think is great, because the glam journal papers (which, granted, have thousands of cites) are mostly cited by people who barely read them. So perhaps you say, well obviously that means the glamour journals have higher impact. Maybe. But I look at similar papers to ours in similarly glamorous journals and they have, say, 2 citations. Or 34 citations. Or whatever, after fighting through review for months. Meanwhile I have preprints and software manuals with more citations than that. But those aren't "for credit" because... shit, I don't even know why not. It's just an administrative quirk. Or something.
STAP and #arseniclife and the recent "MS gene" paper in Neuron provide endless examples of the fallacy that "peer reviewed" is necessarily better ("peer review is boosting with three weak learners"). But maybe that isn't the point. As has been said previously, Deans may not know how to read but at least most of them can count. :-/