Below are a few of my personal experiences (as a student) with this, followed by some (heavily qualified) possible lessons that can be extracted from them.
One of my high school math teachers would do the day's lesson at the board, then give us the remainder of the period to start on the night's homework. People would get going on the homework right away because they knew they could get help immediately if they were getting stuck.
A line of students would form at the teacher's desk as people got into the assignment and needed help. After a while, a second line would also form at my desk. The teacher never even acknowledged that this was happening; neither encouraged it nor interfered, nor even monitored as far as I was able to tell.
As a student, I was often able to understand what another student was confused about when that student asked a question, while the teacher (maybe because of the age/generation difference, or because the teacher was so far temporally from learning it herself that she had forgotten what was confusing about it) misunderstood the student's real question and answered something else.
Some teachers would let me break in and answer the actual question; in other cases, I would have to wait, then pretend to be confused about the same thing but ask it in such a way that the teacher would understand and address the actual source of the confusion.
One other instance I remember is in 8th grade (generally, the school year during which you turn 14, for those unfamiliar with the American system) we had a section of the class that was accelerated. One day we had a substitute who didn't know the material and I taught the lesson to the non-accelerated part of the class since we had seen it a few days previous.
[NOTE: large-ish edit follows, I just realized that some experience from grad school was perhaps more relevant than anything I'd said so far]
Finally, let me relate something from graduate school. I was pursuing a Master's degree in mathematics, and by this time had become reflective enough about education that I started actively observing myself in my interaction with the learning environment. One thing that I observed was that I was never able to keep up with lecture. Another was that mathematics is taught completely backwards from how it is naturally learned (for example: when you are working on a new concept, the first thing that is taught are axioms and definitions; when you are discovering mathematics, you never even think about defining things until you've started to see some significant relationships that you feel need a name, and axioms probably come last). But to return to the point--it hit me very hard one night at a study group that the only natural way for me to learn was to have someone else start asking questions about something that I kind of have some understanding of, at which point I would start to explain it as far as I knew, and then my brain would kick into high gear as it attempted to refine the explanation--at which point I would start actually figuring out the material for real. In theory, this would work if I could convince myself to attempt to explain something to, say, an inanimate object. In practice, the only time I do this is when I see someone struggling with the concept and I start trying to help them.
A proof is in some sense an explanation, and that did work, too. But I had to drag myself to the task of writing a proof, whereas if I could see another human being needing help with it, I would work on it until they dragged me away.
I was one of many people who hit graduate school and suddenly realize that they don't know how to study, because until then nothing had really challenged them. It was a huge breakthrough for me to see that there was a way that I could grapple with the material that was a natural and pleasant experience instead of "why did I think I wanted to pursue a degree in this at all?"-level drudgery.
If there is an overarching generalization to be made about this series of anecdotes with a self-selected mathophile (and obviously there may not be), it might be that peer instruction can happen naturally, if you don't do anything to interfere with it. Another takeaway might be my school-kid observation that the teachers often didn't understand what the kids were asking, while I (another kid) did. This experience was repeated many, many times for me, and not just in math, and I saw other students doing similar things (understanding the asker's question better than the teacher did, I mean). And, the graduate school stuff I mentioned suggest another possibility--there may be kids that only learn when they are explaining. In other words, it's possible that not only is peer instruction appropriate, it's the only optimal way for certain student/subject combinations.
