mk: You might appreciate this recent paper. They looked at a signaling pathway downstream of NMDA receptors and saw that DNA breaks were a natural part of the transcriptional activation of a number of genes associated with learning and memory. Definitely an exciting lead, though it's unclear yet what exact role these breaks and mutations have in aging and disease.
Wow. That is very cool. Just two days ago a friend of mine ask whether or not I thought DNA mutations could be beneficial in a sense outside of evolution. I said that I couldn't think of an example, but wouldn't exclude the possibility. From a purely physical standpoint, biological machinery is agnostic in the realm of possibility. If it can happen, it probably does. If it is reproducible, it is probably of use.
I'd guess the mutations are sort of a side-effect in this scenario, but there is also somatic hypermutation in the case of antibody maturation.