we already have a volume automation curve, which basically multiplies the output volume of a track for a given value each time.
This is good for RNI, but it’s not so good for multilayered samples:
multilayered samples let you play a different sample depending on the velocity value (for example: an hard drum shot at higher velocities, a ghost note (damped and soft) at lower velocities).
When applying a volume automation curve to a multilayered sample, for example for a fadeout, it fades correctly, but the velocity is still the same, and this is probably what you don’t want if you’re using a multilayered sample.
To avoid this, you actually have to put fading volume values directly into the pattern and remove the volume automation curve.
So what about a velocity automation curve which will do it for us?
Damn… I’m getting old and lazy… 
edit:
don’t ask me why this is not in the “Ideas & Suggestions” forum 