That's a very good question. The simplified answer is that the resistor which sets the input impedance is usually sized the same value as the global negative feedback resistor in order to minimize DC offset from a gain mismatch between the two differential input transistors. The value of the gnfb resistor depends on many variables and they are different among different topologies. That's the main reason.
As far as SQ goes, the smaller the resistance the better. This is way over generalized, but resistor noise is proportional to the product of resistance, temperature and frequency bandwidth. So it is desirable to go for a lower resistance in both the input and gnfb resistors. A lower value feedback resistor also allows a smaller value resistor on the feedback shunt, lowering noise further. Noise at the input is carried all the way through to the output. However, overall linearity is more important for sound quality (lower distortion) than resistor noise, hence the tradeoff as to why not all amps have 10K input resistances.
As far as SQ goes, the smaller the resistance the better. This is way over generalized, but resistor noise is proportional to the product of resistance, temperature and frequency bandwidth. So it is desirable to go for a lower resistance in both the input and gnfb resistors. A lower value feedback resistor also allows a smaller value resistor on the feedback shunt, lowering noise further. Noise at the input is carried all the way through to the output. However, overall linearity is more important for sound quality (lower distortion) than resistor noise, hence the tradeoff as to why not all amps have 10K input resistances.