I actually can believe that a short arm would be quicker in switching gears. We're doing microcantilevers at the uni these last few weeks, so this whole cantilever mechanics stuff has been pushed into my head... and yes, lenght of the arm is very significant WRT inertia.

Let's see an analoguous example: the reading head of hard-drives does a similar action as the deraileur arm - it moves back and forth selecting different tracks on the rotating HD surface. The trend has been to decrease the radius of hard drives, ever since they have been created, and increased seek time is one of the reasons.

Is the difference appreciable for a cyclist??? Now that's the critical question! I dunno. I have converted my last geared bike to singlespeed about a week ago, biding goodbye to the last deraileur. Won't be missed.