Negative feedback does not work, in the final aspect of looking for perfection in it, in looking for perfection that is reasonably attainable (compared to other possible solutions).
Positive feedback might, if it is programmed correctly. A mighty big if, if there ever was one.
Problem is, that in an entire playback chain, that positive feedback aspect has to be carefully programmed. But some impossibilities remain, or at least seem impossible. Chaos/’infinite variation’ aspects, at least with our normal level of ability to unwind their complexities.
Negative feedback, one might say, is the simplified method of getting past those potentials in error within executing positive feedback. One that fails to take on the fundamental. Rather that it is ’clever’ and sidesteps it all, instead.
I did do a design where I combined negative feedback with a specifically shaped interference in the given amplified signal. It tends to sound like the best of both worlds. People remarked that they’d never heard anything like it before.
An example of this sort of area of thinking is found in Jim Strickland’s Accoustat TN amplifier circuit. It looks kinda dumb at first glance. The trick is that it is dynamically active. It is transient wave shaping, in it's feedback effects, in the realm of time and level.