What may be more believable, based on pre-existing evidence, is that China’s filtering technology may employ some sort of protocol-sensitive snooping. They can and do watch HTTP GET requests and disrupt TCP connections based on data collected (keywords); doing the same for chat services would be only moderately more difficult. Certainly for mediums such as Yahoo!, disrupting the connection between central chat server and individual client saying undesirable things would be trivial. “Big mama’s” warning users sounds farfetched but is indeed possible if the traffic is being monitored. And this could theoretically be done at any level, either by the chat server itself or somewhere in between client and server. In other words, given what we know about how they filter the web, this is what’s possible and plausible, and it is important, as you say, that researchers, journalists and activists make a clear distinction between possible, plausible, and true, the latter being both dangerous and harmful.

I agree that theorizing about what might be happening is a completely unproductive pursuit, and until there is empirical data on the subject, making claims about the scope and manner of surveillance/filtering of chatrooms is harmful to those of us seeking facts or trying to affect change. Perpetuating untrue/unconfirmed myths or creating new ones dilutes the facts and weakens the anti-censorship position.