There is a Google patent about varying the rankings by time of day, day of the week and part of the year. I've seen this with certain commercial searches in the past and commented on it here.
One example was an EXTREMELY competitive commercial phrase (many hundred millions of daily searches) where one site would rank at #4 between 10am and 3pm local time - and then fall to the second page. This was a consistent pattern for about 8 weeks, bu only on week days.
Why wouldn't Google do something like this and test it? They are not running a "ranking contest" - they are hoping to direct their users to results that match their intention. It certainly seems to me that there could be significant variations in intention by time of day.
If you're used to the old school SEO approach - where you make changes and inch your way up the rankings, tracking what specific changes had what effect - then you may be out of luck today. It still seems to work a little bit like that right now, but not really all that much any more. Google uses statistical sampling, and we are pretty much in the same boat - we've got to look at the big sweep of things.