There seem to be broadly three types of Open Access (OA) journals (and I won't bore you with green, gold and other fancy colours here). In my world, there are journals that are owned by dodgy OA publishers, they likely make the bulk of existing (pardon me, non-existing) OA journals, then there are commercially successful behemoths like PLoS and biomed central, and last but not least there are well-intentioned efforts by academics aiming to break the stranglehold of subscription based journals on their discipline.
There no need to say much about the dodgy publishers, there's a reasonably comprehensible list maintained for that purpose. There is also little to say about PLoS and biomed central. They have managed to get their hands on the pots of cold funders typically empty over STEM subject research grant holders. Typically they charge an article processing fee sometimes approaching thousands of dollars to maintain their infrastructure, and, if they're for-profit, to keep their shareholders happy. There's no equivalent to these sorts of pots of gold in the arts and humanities, yet academic and policy debates about OA are typically driven by STEM folks who are ignorant of the different circumstances faced by academics in the arts and humanities disciplines. Pay-for-play would render many academics unable to publish peer reviewed content in those disciplines.
So, the pay-for-play (aka OA) campaigns quietly forget to tell us who'd give us the money to play under the new OA regimes proposed generously everywhere. Unsurprisingly, research funding councils haven't made up for the in-the-future unnecessary expenditures on journal subscriptions by handing over pots of gold to libraries so that they can fund OA publications. Really what is being shifted here is the responsibility for the financing of research publishing to academic researchers, across disciplines, regardless of the funding situation in those disciplines.
The last group of OA initiatives tries to address this. It relies on academics exploiting themselves in order to maintain the publishing infrastructure that commercial publishers typically maintain (for a steep price). As one would expect of well-intentioned academics they initially offer the having-your-cake-and-eating-it option whereby they maintain the journal infrastructure free of charge, until they eventually fold up or begin charging. The having-your-cake-and-eating-it thingie never tends to work that well in the real world.
A case in point here is the demise of Canada's Open Medicine journal. It was set up as an OA alternative to the Canadian Medical Association Journal after the latter experienced a major scandal involving the publisher interfering with its editor's and editorial board's editorial independence. The journal initially tried to be both OA as well as free of charge, but decided (too late) to begin charging article processing fees. After 8 years and undoubtedly many volunteer hours by its excellent editors, it closed reportedly its door today. The editors note with gratitude the thousands of volunteer hours given to the journals by its supporters. At the risk of coming across as an ungrateful brat gloating over the well-intentioned journal editors' failure, it isn't clear to me at all why academics would want to get into the operational side of the journal publishing enterprise in the first place. As an editor of the equivalent of a monthly journal I can assure you that it takes a lot of time to just deal with the content aspects of journal production.
As far as I can see, this tale is telling. Either you end up with a commercially sustainable business model (PLoS or biomedcentral like), where costs eventually are still incurred at very significant scale, just not by libraries but by authors (in fact, the greater your research output the bigger the financial hit you take!), or you accept that subscription fees will remain a necessity to maintain professional academic journal publishing output.
