I haven't gone back and checked the physics behind what is going on, but it does happen. Heat sinks and/or water cooling works, but they add cost and drop reliability. The real fundamental question is: Do you care about efficiency? As mentioned before, most places which have very high temperatures also have very high solar potential. You can get all the power you want. It is just a matter of cost. I strongly suspect that the right way to solve the problem is to add panels, potentially as many as 25% more. It will be cheaper and more reliable than trying to cool them.
I am actually sure that if you run through the numbers (integrating solar availability over the day as modified by actual hourly air temperatures and expected solar cell temperature rise) you will find that the penalty is far smaller than 25%. It isn't that hard to do, but does take some time and research on your locations diurnal and annual weather patterns. Simpson integration in Excel is more than good enough.
Don't forget to think about things like dust on the panel, and even worse salt water corrosion if you are in a coastal location.