I purchased a MacBook Air 13" (late 2010) about a 6 weeks ago. It came with 10.6.5 installed. I used the 10.6.7 combo updater to bring it up to 10.6.7, followed shortly thereafter by the 10.6.7 supplementary update to address the freezing iTunes problem.
Having no personal pre-10.6.7 benchmark, I assumed the extremely poor performance and experience with Portal under Steam was just the way it was, and wrote it off largely to Valve/Steam. With Portal 2 releasing, I was motivated to dig in and see if there was a way to get better performance. Meanwhile, I purchased Portal 2, and gave it a go. It wasn't even playable, despite dropping resolution, graphics settings to 'Low', etc. Fast-forward.
Last night, I pushed a TimeMachine backup, booted off the USB media, did an in-place install of 10.6.5 (delighted to find it's non-destructive) and then downloaded individual (non-combo) updaters for 10.6.6, 10.6.7 and 10.6.7 Supplemental. After applying them, Portal 2 is running beautifully on my MacBook Air with native resolution and 'Medium' to 'High' settings on all graphics options. The difference is dramatic.
In the end, this reflects very poorly on what clearly can be an admirably high-performing piece of hardware for gaming – despite it's design goals. Chiding the game companies to step-to with updates for Apple's largely mysterious OS update releases misses the point. Clearly, something is amiss when one means of arriving at 10.6.7 causes a dramatic drop in performance versus another method of arriving at the very same point release.