I think people who say "circles will fail" aren't using them in the way most users will. Categorizing people into circles is pretty fun, so if you're a "list person" you can get carried away making ephemeral lists that would change a lot, like "people I'd like to go to the Grand Canyon with." Yeah, that won't work.
But you have to work hard to get lost in the weeds: the product itself suggests four very stable circles: family, friends, acquaintances, and following. Maintaining those categories requires no effort at all.
I don't think critics have thought very hard about those distinctions even though they're so fundamental to the way people live. That's because there hasn't been a service where those distinctions could matter. Many services have lists, that's not what matters. What matters is if on a particular service you would want to say some things to your closest friends and other things to a larger group.
G+ aspires to create and support all kinds of social activity, not just the subspecies of "social activity with undifferentiated masses of people who may want to see my stuff." That's cool, nothing wrong with that. And lots of services have supported that in one way or another already. It's a great thing. But it's not all there is.
G+ may fail to support the other kinds of social activity -- that's always the risk you run when you try something new. But that won't be because it's too hard to identify or maintain different circles. It'll be because when you have a hangout you really don't care who you hang out with. Or when you huddle you don't really care who you converse with, who can make your phone vibrate by sending a message. Or when you check in you don't really care who sees it.
Until critics start incorporating that into their thinking, they'll keep getting it wrong.