Forming a CoI also seems very difficult, unless perhaps you look at it through the facilitation of interactions. Of course interaction along will not do it. I have two examples from work. In two different teams we had somewhat of a CoP but very different results came out of those two groups. In one we openly discusses and brainstormed about a project which lead to new ideas , capabilities and greatly shortened the learning curve. In another the CoP interacted just as much or more but for some reason we became stuck in analysis paralysis. We had interaction in both cases with similar types of people but the context of the interaction and weight placed on it was very differennt. Despite this it seems that a community of practice could go a long way in refining skills, identifying best practices, saving time by counseling together, coming up with better solutions that have been thought through more thoroughly and avoid problems/mistakes.
Something I would hope to learn more about is the idea of how one becomes a part of a CoP and the process of them going from newcomers to experienced members of the group. One idea that I found deals with the idea that a newcomer does simple, peripheral type activities within the community that still contribute to the community's purpose. They continue to increase in activity until they are experienced members. Wikipedia has a little blurp on this. It is called the legitimate peripheral participation theory that describes this process but one of its creators later abandoned or transformed his thinking on this process. I am not sure why.