While the student evaluations may not always be accurate or have merit (and in the age of professor rating sites on the net, one wonders), I think there is very little usefulness in peer evaluation. Most of the time, we tend to select folks reciprocally. I'll ask a friend to come to my classroom; I'll go to his. The unspoken quid pro quo is obvious: we each write glowing summaries of the other's teaching. These are folks with whom I will spend the rest of my career, and I haven't the slightest intention of putting competence before collegiality. That sounds irresponsible, but honestly, the irresponsibility is within the system itself.
This is not to say I don't ever criticize my colleagues. I once had a student approach me about a faculty member whom she felt was harassing her; I did indeed go and have a sit-down talk with him at once. Where student safety is concerned, I'm not afraid to get in anyone's face. But when it comes to teaching methodologies, lecture strategies, and syllabi choices -- I prefer to"let my colleagues be" because, by gum, I want them to"let me be" in return.
Even our division dean is part of this. After all, division deans are faculty members too, selected from within the department. If they anger tenured faculty, they are removed from administration rather rapidly. Though they can afford to be candid with adjuncts and the untenured, wise administrators ignore all but the most flagrant cases of incompetence in the ranks of the permanently employed.
That being said, in the end I suppose that student evaluations (informally on the 'net or formally in the classroom) are more likely to be honest reflections of teacher performance than any other instrument.