> No more than the background noise, it's tough to judge with a Web Server. I > use -xO4 in all my test builds so it seemed natural to use it here. I'll > revert it to -xO3
I agree with Jyri here - what have you been using from a testing perspective? If it's been with -xO4, then I would suggest using that unless you've encountered failures or regressions of some sort.
