At a glance:
While the arXiv does contain some dubious e-prints, such as those claiming to refute famous theorems or proving famous conjectures such as Fermat's last theorem using only high-school mathematics, they are "surprisingly rare".[better source needed]
A moderation team and "endorsements" are great, but for most of the journals that these large distributors operate, there exists a system. The author submits an article, and several well-known folks in the field review it extensively (hopefully). In almost all cases, the reviewers are payed to give constructive criticism. Occasionally they will even recommend that the article undergo major revision before publishing. Either way, the current methodology has a bureaucracy that ties into major publishing firms. This made more sense when we only had physically printed paper to distribute articles, but I think a lot of scientists are looking to cut out as many middlemen as possible, so long as the legitimacy (or apparent legitimacy) of their work is maintained.
So now, open access journals exist, such that instead of incurring cost to the reader, the author pays for the peer review system with the hope of increased distribution, as there is no longer a paywall to read their results. But (b_b touched on this) how do we remove the prestige of saying "I'm up to 3 articles in Geophysical Review Letters!" (or whatever)? Some journals have career-making reputations. I think it will take time, but the prestige will see redistribution.
This article is three years old (and more than a bit ironic), but if you really want to dig deeper, it's got the nitty gritty.
Edit: Also, the golden nanorods thing wasn't necessarily a case of poor peer review, it was a case of complete data falsification. You can't immediately protect against things of this nature, but reproducibility will eventually out the liars. Well, the experimentalists, at least. Theorists are another story.
Edit2: Good peer review will probably never be free.