Lindrone - This isn't a response to your post (math exams on Wednesday and I ain't got time to think that hard!), but I wanted to throw this thought out there: The decay section of a note is (almost always - barring Paul Revere-esque trickery) of a much longer duration than the attack phase - making it much easier to discern. Screw up something which lasts for .75 seconds and it'll be much easier to tell than the .18 attack phase.
At which point I think it's safe to say that while it may be indeed possible to modify one portion of the note and not the rest of it, it's probably much more likely that we're not as attuned to that phase and see the effect as being only to the decay.
I suspect there's a more scientific answer at the root of your theory. The very high frequency data (>12k) probably plays a big role in it. I have a hard time believing that if you recorded a reproduced sound of each high-end 'phone, the actual duration of the sound would be too terribly different. [Oops - didn't mean to introduce empiricism here, and ruin the whole thread.] Models like the Etys do give an impression of being more incisive in their presentation of short bursts of sound, but I still doubt they could sit down and say, "let's shorten the decay phase of the note" and make it happen.
Occam would probably shave my legs with this one, but I think it's too simple a theory to be the complete answer. There may certainly be an element of it, but I still maintain that as long as an electrical signal (i.e., audio data) is still being pumped into the driver, it should move. As soon as the signal stops, the driver should stop, and should do so with a linearity to match the signal. If a driver is doing it more slowly, that could be easily explained through inertial phenomena - if the driver stops before the signal (and/or at a faster rate than the signal), that would imply a stiffness to the driver which would surely seem to manifest itself elsewhere. E.g., putting your hand on a speaker cone kills decay - it also kills all the nuanced movement, therefore higher frequencies are muted.
Ehh...I guess that was a response.
A speaker doesn't know what's attack and what's decay - especially considering complex waves, which is what it's constantly reproducing, unless you're listening to a test tone CD.
I'll e-mail my super-duper-scientific contact and see what his impressions on the issue are. It just seems counterintuitive to think that you could construct a driver to artifically shorten the decay more than the source material (without electronics) without making huge compromises elsewhere.