Interesting discussion here!
As for the dry fire speed: Steve is right that arrow speed continues to accelerate as arrow mass decreases until zero, but it will eventually reach an asymptote (it will level off): speed increases will gradually become marginal.
The last comment by Willie about usable accelerations during the power stroke made me think about how Force-draw curves might influence this.
For example, a hump-shaped FD-curve means that you add less energy in the last few inches of the power stroke, and hence during release the acceleration of the arrow is small at first. With compound bows with asymmetrical cams this is extreme: the max force is reached halfway the draw, not at the end like with normal bows. You can actually also see this effect in high-speed recordings of arrow release on wheely bows: at first its acceleration is low, and in increases steadily.
Unfortunately, we need to take the dynamic (instead of the static) force-draw curve into account, which is the FD-curve upon arrow release, and which changes with arrow weight and spine.
In straight bows (with straight FD-curves), max acceleration with very light arrows is achieved shortly after arrow release. What I wonder about, is how do stacking bows affect speed of very light arrows? I can imagine that stacking doesn't matter all that much when the only thing that counts is arrow speed.
