There has been many good comments about the recent opinion pieces on PLoS from Nature employees (you can find the link via the comments - here is the second).
One perspective which has been overlooked a bit, as far as I can tell is that Nature has managed to frame the debate. Intentionally or not.
The framing can be summarized as "the problem is that top tier journals can't be profitable as open access - author pays, in competition with Toll-Access journals". This framing is advantageous to Nature as it implies that "the author-pays principle must generate all revenue for OA journals - subsidies are not ok". Equally advantageous is it when "top-tier journal" is defined not only as "high rejection-rates" but also "high overheads". Nature has an interest in making the two things seem inseparable.
For society and everyone who is not Nature (or other Toll-Access, high overhead journals) the framing does not make sense. The debate is part of a broader debate of the future of scientific publishing. And it is unreasonable to assume that a future of efficient digital publishing must be hobbled to serve to needs of businesses adapted to the past of high cost of paper distribution. Or that it must be measured by the same criteria of success (high profit from monopoly priviledges) as old businesses.
For society the question (obviously) is "how do we maximize society-gain from published research". There is no question that the enormous subsidy that is coorporation held monopoly priviledges via copyright can make publishing of popular content profitable for some publishers. But since this arrangement maximizes monopoly profit for select organizations and not gain for society, society should look at other arrangements. Other arrangements, which secure access also for those that value the access below the monopoly pricing but above the neglible cost of digital reproduction. And in no area of publishing is the fairness of society interference with current buisiness models larger than science, where the producers of content are mostly publically financed and the consumers also.
Among "other arrangements", one obvious extreme is lump subsidy and no access restrictions. This gives maximum society benefit at a fixed price.
Other arrangements are "some subsidy + other revenue" which is exactly the model that PLoS explores. And which should be explored. And which does not need to be successful by the same measures as Nature was successful in the past.
And as a personal opinion on this - I think that what will (and should) happen is that the market (lead by initiatives like PLoS) will drive costs of top-tier publishing down. Future top-tier publications will simply have to have unbundle, have less costly content, and less costly procedures - and will consequently look different from today. I see an obvious analogy to newspaper-publishing which was discussed at the Becker-Posner blog recently: Posner - Becker
And as Greg Laden relates, Nature supplied somewhat opaque answers to the UK parliament on Nature finances. But from those, to me, there seems to be plenty of costs to cut.
(Updated: fix link, clarify last sentence)