DPA has specs on the distortion of their mics. I'm sure other serious mic companies do too. The distortion spec is similar to how people spec power amps: this amount of sound pressure level will cause this much distortion. For example, the 1% THD spec (or -40 dB) for their 4006 omni capsule is 136 dB SPL RMS, or 139 dB SPL peak. Its max input SPL is 146 dB SPL which causes 10% THD (-20 dB). They wrote a whitepaper on it too:
the-basics-about-distortion-in-mics
A cool factoid from that link: air itself goes non-linear (ie. distorts) at 194 dB SPL because you can't go below a vacuum for the negative part of the soundwave.
For feedback amount and distortion, consider that feedback feeds the whole signal back into the input, so whatever distortion is present is compounded again, which has two implications:
1. feedback can cause more distortion products, albeit at a much lower level
2. the better the open-loop performance of the amp, ie. the less it needs feedback to get good numbers, the better it will probably sound due to 1.
The mathematical basis for 1 is easy to see if you do a power series expansion of an amp's transfer function. A perfectly linear amp looks like f(x) = a*x, where the input x is multiplied by some constant a to make the output f. A non-linear amp (in both the mathematical and the engineering sense) is one where there are higher order components: f(x) = a*x + b*x^2. In this case, this amp will generate 2nd order distortion because it has a the x^2 term. The exponent (2 in this example) is equal to the order of distortion it will make.
Now consider 2 of these devices hooked up in a row. The output will be f(f(x)) = a*f(x) + b*(f(x)^2). If you work out the math in this, you will have x^3 and x^4 terms, which means the combo of the two devices is making 3rd and 4th order distortion in addition to 2nd order distortion. So the distortion compounds. Each stage in an amp will compound their distortion products too, and that might explain why some designers eschew very complicated, multistage amps.
In the case of negative feedback, it's kind of like this except the output is fed back at a much lower level and somehow subtracted from the input to make the output come as close as possible to the input, but it's still generating higher order harmonics which are not subtracted. And higher order harmonics are bad because they occur further away from the fundamental tone (the input signal) and are therefore easier to hear – it's more likely to be outside of the critical band of the ear at the fundamental tone's frequency. So that's the theoretical basis of more negative feedback potentially causing more audible issues in an amp, and why the cleaner the open loop performance of the amp, the better feedback might work for it from a subjective POV.
You can also see why global feedback might be subjectively worse than local, or single-stage feedback, because the global feedback will have the compounded distortion products of many stages of an amp in it whereas a single-stage feedback will only have its own stage's distortion products.
Note that this is not a novel idea. I think I read about this in some AES paper from the 50s or 60s.