And PLOS will waive their fee, but in practice that's rare. If all the people currently paying little/nothing to submit to non-open access journals instead submitted to PLOS and waived the fee, PLOS would be fucked and they'd have to change their policy. After all, they have to pay a lot of salaries: http://www.plos.org/about/people/staff/
http://www.earlham.edu/~peters/fos/newsletter/11-02-06.htm#n...
Note also the linked studies. Summary: the majority of OA journals don't charge fees.
"in practice [PLoS fee waivers are] rare": well, among the people I've spoken with, 100% who asked for a waiver got one. So at present it's apparently not difficult to get a waiver. And when you say "they'd have to change their policy" that simply does not follow. What is true is that they'd have to figure out some way of funding it. Eliminating fee waivers is just one approach. But there are many other ways it could be funded. What would matter in that instance is how committed the people running PLoS are to the waiver.
There are lots of examples of various people in other fields (music, indie video games, restaurants, etc) charging a 'pay what you can' strategy. But not many institutional ones.