> noise seems to lower with the higher value resistors in tube microphones.
It's complicated. A lot of interaction. Over most of the audio band, the capsule capacitance is lower-Z than the grid resistance, so it shorts-out the resistor noise, and a larger resistor is shorted-out better and the result is lower noise.
But that happens when resistor noise is fixed-power thermal noise. Double the resistance means 1.4 times the noise voltage, but 2:1 better attenuation from capacitance, so 1.4 times less noise voltage.
When the noise source is a pure current, like grid shot-noise, double the resistance is double the noise voltage. Also double the attenuation from capacitance. So with a tube, increasing the resistance may not change the midband noise much. (Noise current in the grid/gate of a small FET is 10X to 100X smaller than tube grid current, so FETs can use ultra-high grid resistors.)
But the situation changes in the bass, where capacitive impedance is no longer much-lower than resistance. The curves bend. And what is best in midrange is worst in deep bass. And the curves generally slope somewhat like Fletcher-Munson, so we can't just say that bass noise is inaudible. So the "offensiveness" of the total noise versus resistance may be very much a matter of taste (and how bass-noisy your studio is).