My two cents: As it stands now you can easily use twisted if you wish and it doesn't really need to be a part of pygame proper (an inference I am perhaps imagining).
It seems in the interests of pygame that you would provide a standard interface for networking and then developers would write implimentations for the various network transports avaliable. I've used twisted a fair bit and my only issues with it would be: the twisted documentation, integration of its reactor pattern into projects and the overall size of the twisted project (to be considered when distribution and maintenance issues come up). The advantages I see of a using someone else's library (pyraknet, twisted, etc.) is that the code wont have to be maintained by the pygame team and most of the work is done. The disadvantages are that pygame may become dependent on yet another third party (it seems at this point compromises are already being considered for surfarray, though I may be wrong here). Also, I don't know how many pygame developers depend upon py2exe/py2app but this may be important to consider when choosing a specific transport technology. Regards, Jotham