Then you try to cancel that all out with the feedback
That's NOT how feedback works (in some sequential manner). You should learn
something about proper feedback theory.
Well, there is some propagation delay of the signal as it works its way
thru a real op-amp. Music signals are full of transient spikes and
such, not just continuous sine waves. This propagation delay doesn't
seem to be specified or even presented in most datasheets. But looking
at a somewhat randomly chosen op-amp
http://focus.ti.com/lit/ds/symlink/opa244.pdf
on page 9 there's an oscilloscope "picture" of the "small signal step
response". The op-amp is set to unity gain, the output has an overshoot
of about 2uSec duration. That would be the period of time the signal
took to get from input to output, and thus how long it took for the
feedback loop to see it and apply it to the other input of the op-amp.
And to correct the overshoot. Sure, that's not much time, about a half
cycle of 250KHz, and okay, that's probably fast enough of a response for
audio work. But if it was say 200uSec, that would be a period of time
for a cycle of 5KHz and a period of time that the feedback loop isn't
doing anything, and you'd have audible overshoots, and it would sound
pretty bad. Fortunately most op-amps are way better than that, but I
mention this excessive case to demo this point. Of course I may be
worrying about something that doesn't normally matter, and thus I may
well be full of Total and ABSOLUTE ********... :-)
Sure, if you had an amp that had zero propagation delay, then feedback
would be immediate and what I mentioned about the feedback loop having
to clean up secondary crud would go away.
You get a similar effect with vacuum tube audio amps with the global
feedback loop. There is a time delay (aka propagation delay) for the
signal to get from input to output. This is also expressed as phase
shift, and it gets bigger for higher frequencies. There's some high
audio frequency where this phase shift hits 180 degrees. And that makes
normally negative feedback go positive, and it could go into
oscillation, so you build the feedback circuit in such a way that it
gives no feedback up at such high frequencies.