Hi again,
I was going to say, if the feedback is too fast then it causes audio distortion, so there is actually a limit to what the speed of response can be, and it's usually set with an RC network.
For example, if the audio coming in is 100Hz and the time constant is only 0.01 seconds, significant sine distortion would result. Thus it's a tradeoff of compression speed vs distortion and usually the distortion takes precedence. The drawback is the 'breathing' effect.
Using digital signal processing this problem can be alleviated to some degree.
A feed forward system is very simple to analyze, so i'll try to get back here with an example some time.