I've written blog posts, several meta posts, and said many other times that "It is simply too early to apply numbers to the process." It's easy to look at a site and say "yeah, that site is working," but what does that mean, exactly?
Q&A sites tend to run heavily on network effects. If you have "X" amount of activity, the site runs pretty well. At 90% of X, the site will start to spiral downward. So what is "X?"
I guess we can make something up, but we'll likely get it wrong and kill many good sites in the process. Or, we can wait until we have real data from live Stack Exchange sites to number crunch -- to determine at what point does the site "work" and when does it just flounder.
A rough definition of a successful site is that most questions receive an answer with a reasonable amount of time. So, what's "most questions?" What's a "reasonable amount of time?" Some sites will be really large. Other sites will be very successful with a smaller audience. Some questions can be answered in seconds. On a highly theoretical, niche site, it might be reasonable to wait a week or more.
All of these issues have to be taken into account when deciding the "criteria to decide if a Stack Exchange site should survive the beta or should be dismissed?"