Something I learned from Marc Coenders is to make a distinction between communities of practice as a (socially-constructed) learning theory and the 'thing' we call communities of practice. This sounds really logical and evident but helped me to be clearer at times what we are talking about.
The community of practice theory talks about how people learn, and how this learning is linked to being part of a community (socially constructed), working from empirical observations on how learning of practitioners actually takes place. By observing and understanding what happens 'naturally', we may be able to be better facilitators, advisors, coaches, members or leaders of learning processes.
On the other hand there are 'things' (usually groups of people- but sometimes mistakenly used for platforms) called communities of practice for whatever reason. Because the initiators were inspired by the theory, because it sounds nice, or because someone in the organisation heard about the term and introduced (and it stuck). These 'things' may also be called differently, for instance learning circles, and in that case the theory can still be helpful to make sense of what's happening and to see how best to intervene to make the group effective towards its learning purposes (this may not ring any bell to you, but it helped me, so I thought I'd write it down).