If you have good development practices in place, making a targeted build to prototype 3px UI improvements suggested by ANY trusted team member should be trivial and should be considered. After prototyping the change, it's possible to evaluate whether the suggested lightweight improvement falls into the range of essentially invisible (probably not worth the churn late-game, but fine earlier) to actually awesome (crank a build if you can). Even lightweight changes could break the hell out of some browsers/resolutions, so 3px is not really "free". Hell, 1px is not free if floating elements can cause layout changes or fixed elements can crash into adjoining ones. If you have challenges making and evaluating suggested changes on very rapid timescales, ANYpx is irrelevant in the scope of your other problems and you should trend against change where possible. Also note that I said "trusted" team member - if the team member pushing for design changes is not "trusted" that is a separate consideration - please weight appropriately. If you break the crap out of other areas, yeah, try the 3px at that point if it's a toss-up and the change is orthogonal to other changes. If you pay the price of looking at other changes, try to get the designer's suggested change tested for free. Trust, but verify.