I'm not nearly as inclined to research this topic and make myself a sound engineer at the end of the day.... but let me just try to throw some common-sense on this topic.
You are talking 2 and 3ms. That's .002 and .003 seconds. I doubt seriously the ear could ever discern such a difference.
Even if it was possible to discern it, making it "audible," it still takes a HUGE jump to say this audible difference is significant enough to translate into the ability to affect perceived "quality."
I would bet there is nobody that could listen to two tracks (one we would use as a reference, and the other we would use as the "delayed" subject), and sit there and honestly tell me one sounds "worse" than the other. I think given what folks are trying to push in this theory, a more "pure" test might be to have it relate to a single note. With perfectly calibrated instruments, record an "A" for 5 seconds or so. Then, place a 2 or 3 ms gap at the begining of a second "A" and dub it right over the old "A" for the full 5 seconds or so.
What you would have in the end is what we are all discussing. You have the frequencies coming in at different intervals. BUT these are the same frequency, so you've removed all the other variables and got right down to delay ONLY. If you did this, could you not measure the pitch with a pitch meter?
I would bet without a pitch meter (just using the plain old ears), there is nobody that could distinguish any change in the pitch (or any so-called "noise" this delay might create). I would also bet a pitch meter (I hope I'm using the right name for the instrument - those things that help people tune their guitars, etc.), would also register a perfect "A."
Just because you can "picture" the wave "noise" that results from time delay in your head on a graph does not mean you can hear it.
I would even bet you could play the single "A" note by itself and insert 2 - 3 ms blank intervals. For example, the note sustains for 1 second, then is shut off for 3 ms, then back on for 1 second, etc. I bet everyone would agree once played back that there is absolutely no discerning any gap anywhere. It would sound like a constant note.
Imagine a woodpecker sitting on a tree and pecking it 500 times in 1 second (that would be 2 ms intervals) or 333 times in 1 second (which would be 3 ms intervals). It would have to sound like 1 "smack" to everyone.
Now, imagine the woodpecker had 2 differently-tuned pieces of wood to peck, and could peck each 1 at a time, at the same 3 ms rate of speed. There, you might hear a different note. But you will not hear both notes. This is what somebody referred to as delay affecting timbre (not timber). But remember, there ARE 2 different notes being played when this occurs; the deal is that they are so close together, they sound like 1 - but only a different 1. That's the same thing that would occur if a guitarist could play different notes that quickly back-and-forth.
However, the music we listen to is nowhere close to this example. Nobody listens to 2 notes and calls them a song (except Ted Nugent). Really, though, when you have such a limited number of notes in the mix, it "could" affect timbre, but it would still sound so "pure" there would be no quality issue.
When you get a bunch of notes moving around in harmony, there is NO WAY you could ever process this immensely minute difference in timbre. The brain is too busy hearing the song.
So, my belief is this delay phenomenon is interesting in terms of picturing graphs and pondering how the "obvious picture" might register in terms of the quality of the audible sound. But I think our brains process sound based upon what the focus of attention is. Ever been watching a news story, and your wife says something, but it doesn't register clearly? Ever been talking to your wife while a news story is on, and the news story did not register clearly? These are the EXACT same sounds, only you heard them qualitatively differently.
If you are being a "good" audiophile - not thinking about anything else but that song, there are still so many notes flying around and so much harmony put together that you could never discern any qualitive difference between frequencies that came in delayed by 3 ms.
By the way, it is only when these delayed pitches mix that we encounter a "noise" experience. Noise is just a perception of one or more sounds that do not "fit in."
What is the frequency of an "A?" Obviously it is in waves/sec. Graph the "A" and graph another "A" on top of the old "A" but adjusted to the right by 3/1000 of a second. The intersections create a new tone. You have a new amplitude and wavelength. What are they? I'll bet the new amplitude and wavelength is so utterly close to the original "A" it's not funny. Compare (1) the variation in frequency of the "new" note from the original "A" and (2) the variation in frequency of the "new note" from the "B flat" that follows up the line from our old "A." I bet the variance shows the "new" note is real snug up against the old "A" and miles apart (relatively) from the "B flat."
It would be like taking a trip to the grocery store versus trying to travel outside our galaxy.
Y'up. I'm sceptical of this theory. If you want to talk 50ms delay... that's different.