Doug, the Intel folks seem to want to have it both ways. You are right - if the power curve is linear then Dothan has problems. If the power curve is not linear then Dothan could do fine while Prescott continues to cook. I have not heard anything magical about rev D so I'm not expecting miracles there.
From Pravin's reply I see that my surface understanding of this topic looks correct. Further evidence is that Intel is changing their whole x86 strategy to be based around Dothan derivatives, the end of NetBurst. Tells me that Intel considers the whole future of NetBurst concept to be a bust - undoubtably because of stubborn leakage/power problems.
Meanwhile, rumors on Dothan's power consumption seems to be all over the map. Believe nothing.