Well, if there are any performance regressions in a snapshot release
then it's better to find out immediately than when we're trying to put a
real release out.

(in fact, I think it's quite likely that a constant factor performance
regression discovered once the RC phase has started wouldn't be fixed
before the release).

I most definitely agree. I just don't have the cycles to do these things very often (as it's quite a manual process right now). The idea is to eventually automated the nofib runs to run nightly. We're not there yet I'm afraid. :/

-- Johan