Math + Logic + Science = something completely mad...

I still like mogami for RCA though they are on the expensive side I think they're constructed very well and sweetwater is great to do business with.

We're on the same page regarding Sweetwater.  Just bought a 7-string multiscale axe from them a few weeks ago, been doing business with them for years.  

Those amplifiers had lots of feedback, but very poor gain bandwidth product. That was why they had bad distortion products, and why they had poor output impedance at high frequencies.

They had excellent THD and IMD.   Designers were sure they have to sound great.  In 1969 I almost bought inexpensive 100W amplifier modules with incredibly low THD and IMD.  NFB increases bandwidth and reduces output impedance.  40dB of negative feedback will reduce output impedance 100 times.

Even when RF gets in, short of causing oscillation it may not even do anything.

You don't even need nonlinear element, like diode, to demodulate.  It is enough if amplifier's slew rate for positive and negative signal is different to demodulate, for instance, AM radio.  It is called "rectification phenomena" and is very common to opamps.  

Yes, high impedance nodes are sensitive, but output of an amp, in spite of lower impedance (many ohms at high frequencies), is an input to the amplifier with 100x higher gain (for 40dB NFB) than amplifier's normal input.  We want to reduce phase delay (improper summing of harmonics) at high frequencies caused by limited bandwidth.  For that we increase bandwidth to more than 100kHz  (my AHB2 has 200kHz@-3dB)  getting very close to AM radio stations, not to mention switching frequencies of SMPS present everywhere (computers, TV) and even LED bulbs.

I am wrong about 100x higher gain, since only fraction is fed back to the summing junction, but it is in the same level as nominal amplifier’s input. It won’t be as sensitive because of lower impedance, but will still inject noise picked up by long unshielded speaker cable.

Negative feedback increases the bandwidth w.r.t. the definition of bandwidth as the -3db point. However, it does not increase gain at that frequency. We are both going off memory but I am almost certain an amplifier in 1969 didn’t have the IMD numbers you are quoting. I wouldn’t guarantee it didn’t but I highly doubt it. Can you please confirm that in some manner?

No, I remember year and even the store (spending vacation in London) but don’t remember model number or brand. You are right about lack of excessive gain at the high frequencies but I’m not sure how bad amplifiers in 70s were.

Rectification phenomena is not even specific to electrical processes. It applies to thermal processes as well and can cause non linear electrical effects modulating back into the electrical signal, typically low frequency, but you can actually detect it in switching MOSFETs.

Rectification phenomena, as it was described by Analog Devices, at their seminars is an electrical process. Uneven positive/negative slew rates combined with limited bandwidth result in DC proportional to amplitude of high frequency signal (hence demodulation) - at least that’s what I remember. You will easily find few opamps advertised as rectification phenomena free. I’m not sure if slew rate difference is the only reason for rectification, but you can find more here:
https://www.analog.com/media/en/training-seminars/tutorials/MT-096.pdf

The emitted RFI from switching supplies for consumer electronics is fairly low and the efficiency of conduction into a speaker cable would be low. You are correct the feedback gain is high, but still usually bandwidth limited, and while there are RFI noise sources, the efficiency of local wiring and audio wiring for picking up RFI from an antenna standpoint.

It depends on switching supply. Some of them, like resonant mode SMPS, are extremely quiet while crude high current computer supplies pollute everything around. Amplifiers are bandwidth limited (not limiting  already rectified/demodulated signal), and antena (speaker cable) has very low gain for <1/10 of the wavelength. It means that interference will be greatly reduced but not eliminated. I assume we can hear -60dB down from nominal signal level (equivalent to 1mV signal level). As for the standards - you can find them for everything.
Bearingless torquemeters (that I used to design electronics for) digitally communicate from rotor to stator by high frequency pulses and two sets of coils. We had to lower frequency from 20MHz to 9MHz because field intensity was violating standard. Our competition almost lost business (had to stop production) because of that (Navy complained). In the proces we purchased NARDA calibrated EMF measurement device (scope).