I see this as highly problematic from a DIY point-of-view and to a lesser degree (all puns intentional) from an engineering POV, but I certainly have no first hand experience. The problems I see are:
1) How do you sink the roughly .85kW/m^2 of heat effectively w/o adding tremendous cost? Obviously you envisioned using a liquid coolant, but how do you transfer the heat to the liquid? Certainly a hollow aluminum or copper backing would do the trick, but that is a non-trivial additional cost.
2) Assuming a cheap enough heat transfer solution could be found, removing the heat from the solar cells is only the first part of the process, you still need to remove the heat from the cooling media somewhere in the cycle. Adding a radiator and a fan would eat up much of any anticipated gain in a small system.
3) Cost to Benefit ratio seems skewed to me. If we take a 2kW rated array and de-rate 20% (that's just off the top of my head, I have NO IDEA what the actual de-rating factor is) due to heat, this is a loss of 400W. It is unlikely that any cooling system could regain more than 50% of the de-rating, so we are haggling over 200W. @ $10 per installed Watt this has a maximum value of $2000, but because it is only truly a problem in the Summer, we should only count the de-rating 25% of the time. This brings us to $500 to spend on cooling a $20,000 array. If we simply added another 50W to the array, we would cut our "losses" in the summer by ~20%, but in the other three months we should see an average of a 2.5% increase in system output.
The other way to look at this is that in the Summer the longer hours of insolation are built-in compensation for the de-rating.
Finally, on the Cost to Benefit venue, If you size your system based on the actual output rather than the name plate output, you can simply brag about the "free extra power you get in the Winter." :-)
If you want to actually give it a go, consider aluminum "roof flashing" bonded directly to the back of the panels. Devise a punch and die to make "cooling fins" prior to attaching. This passive system might help lower temperatures >> than ambient with little actual cost, but I doubt that you are going to see a large improvement.
Keep Thinking :-)