ATI powermodules technically as well... the air inlets are at the ends, so the air really just blows across the ends of the bulbs, leaving the bulb to heat up a bit more. Then the air blows across the ballasts and up through the blowholes in the center.
Other designs like Aquatinics and Current use fans that blow from one end to the other. The technology may not call for that though. If anything, we can hope that the cooling air that is drawn in at one end happens to be the end where we put all our 'cold spots' and that the heat from this is enough, and the air throughput rate is low enough that the rest of the bulbs stay a bit warmer.
The DIY fixtures I made (red house) are ducted like the ATI as well... so far so good. The idea of attaching a heat-pipe, sink, or waterblock just to the cold-spot of the bulbs is interesting though. I almost wonder if just bonding some radial (rounded 5/8" inside curve like a collar) to the end while lowering the air flow might be what the doctor ordered. Heck, perhaps just putting thermal paste on between the cold-spot on the bulb and the reflector clip (retro reflector) might be similar (and the airflow on the backside of the reflector cools the heat conducted to it).
I just thought I would point out that its not as effective to cool the middle of the bulbs as it is to aim at the end with the 'cold spot'. When I did my 6x39wattT5 Tek over the 40B, I hung the fixture so I could place a 50 cfm 12" long cross-flow fan on the end to blow on the bulbs. I had 'dumb luck' with this because the end I put the fans on happened to be the 'cold spot' end where I had lined up all the writing on the bulbs (anal, Ill admit). So my fan was blowing across the cold spots first, and then somewhat along the rest of the bulbs (but not as intense, and with air that was warmer as it went further away). So in the end, I ended up with a nice 20% boost in PAR from this, and I was thinking I should add another fan to the other side to even things out, but upon testing the bulbs, I couldnt figure out why the opposite end of the bulb had just as high an output as the end that was getting more cooling. Well, now I know why... I was cooling the electrode end, or the 'cold spot' which is what matters most.