So small sites will still fail, in fact they'll fail often, but the meta-media as a whole can continue. If anything, it forces them to be responsive because unhappy members will vote with their feet.
It also gets around many of the censorship problems that large sites have. Large sites don't want to make their algorithms or policies clear because there's a massive gain in gaming their system. A diverse collection of small sites is effectively immune to this: it's easy to game one site, but it doesn't get you much of anything.
There's still the expense of moderation, and I think it's where a meta-system shines. If you can draw moderators from anywhere, they're naturally disinterested. You can set up a standard set of procedures for how moderation is done without actually specifying what the rules are. The site owner has sole discretion as to what the rules are, and can override moderator decisions, limited by their willingness to micromanage and upset membership.
The incentive to moderate is that if you want to appeal a decision or otherwise request moderator attention, you must yourself do moderator duty to earn enough "civic duty" points. It's like jury duty, only more directly tit for tat. And I think a well designed system would help educate people as to why certain decisions are made, leading to greater investment.
I think there are huge technical barriers to such a scheme, especially building a UX that works across systems and dealing with not terribly technical site admins. Hell, I'm not even sure what the "meta-media" part of it is, whether it's discussions or interest groups or are there meta-sites that somehow overlay multiple real sites on common topics.
But it would be a fun project...
* Is there a more generic term than "social media" that would better include SO?