I think the ducking is really unnecessary and makes it difficult to enjoy.
I listened to Eegah's sample, and I don't like it. What about when the riffers start talking over the narrator? You need to have ducking at that point, otherwise it's just a mish-mash of talking and it sounds like noise. Is that really what you guys would prefer?
Perhaps the attack of the ducking could be adjusted - I think a reasonable comparison would be to alter what we're doing now by increasing the riffer audio so it's a little louder than the source, and easing up a bit on the ducking. That way it's more subtle. But I think it's definitely a requirement, especially in the event of someone talking over dialogue within the short itself.
EDIT: After reviewing some more of the clip, I do agree that the levels have some quality lacking. It seems like there is a lot of variance in the riffer commentary...
A little bit of ducking is probably optimal (like MST3K-levels), but I prefer no ducking to the large amounts of ducking that are on the clips provided. The ducking for "Down and Out" almost completely wipes out the original audio and, to make matters worse, the riff audio seems to be lower than the original audio (it should be at the same level or higher). Another thing I don't like about ducking in general is when it telegraphs the joke. If the ducking is subtle enough, though, that doesn't happen.
In the end, I'll take these clips however you provide them and like it. Everyone has their different preference and if most people like them they way they currently are, so be it. If that's the case, though, it would be nice to be able to get the mp3's and original source video to make them myself.