Often I find reviews are useful precisely because of the distance the reviewer has to the code; a fresh set of eyes is good to catch issues. (Frequently I'd go over my own changes before submitting for review for the same reason).

Pair/group programming also often ends up lacking in notes on what the thing was actually meant to be doing (because this was communicated verbally and never written down in reviews), which makes code archeology difficult. Or maybe that's just derived from the (external) code base I had to deal with that was done that way and doesn't generalize…