On the Liblicense list, Heather Morrison addresses 'The Religion of Peer Review' and refers to an article by Alison McCook: Is Peer Review Broken? The Scientist, 20:2 (February 2006), page 26.
To ask if peer-review works is probably asking the wrong question. It's a ritual, not a scientific method. It's a cultural expectation. Just like wearing a necktie is in certain circles, and nobody asks whether they actually work. (They would, as a noose.) And to expect peer-review to act as an almost infallible filter is wholly unrealistic. If it is a filter of sorts, it is one that helps journal editors to maintain their journals' biases. If peer-review were a method of only ascertaining an article's scientific validity, we would neither need, nor have, so many journals. One in every discipline would suffice. But the ritual reaffirms bias. The bias of 'quality', for instance, or 'relevance' (though the question could be asked to what, exactly?). And why not? Just as bio-diversity is a good thing, 'publi-diversity' may be as well.
She also asks, in the same posting, if there is "scientific proof that current methods [of publishing] will work?", saying that the "...current approach has [...] led to the serials crisis." She has a point, asking about proof, as the question is being asked of open access publishing, so why not of traditional publishing. But talking about rituals, isn't it a ritual, too, to complain about prices increasing faster than library budgets? Nothing remotely scientific about it. There would be a point if library budgets had broadly stayed in line with research spending. But they haven't. Isn't it an article of faith that the budgets "could not conceivably rise" in line with the production of scientific literature?
Open access publishing, in addition to all the other benefits it has, also keeps the cost of scientific literature in line with research spending. This isn't, of course, proven yet, let alone scientifically. But how would one prove it without doing it in the first place? The proof of this pudding, I'm afraid, can only be in the eating, as the saying goes.