Now that I’ve worked in a team that really was doing XP (except for our Lack Of Onsite Customer shortcoming) I can say that it works pretty well, but only to the extent that you do all of the practices together. Of course, that’s what the XP folks said from the beginning: you can’t just apply 1% of it and make a judgement about it.
With the recent resurgence of the XP methodology in the Ruby on Rails community (particularly in companies affiliated with Pivotal Labs), I expect a whole new flurry of related controversy. Hopefully with the experiences and advice I’ve given, this time we can avoid some of the confusion and skepticism, and actually adopt it where it makes sense.