What effect does that have?
It reduces the amount of duty-cycle modulation that occurs when overdriving the phase inverter. Typically, values in the range of 330k to 390k give the best results, but 470k may be enough. It will not affect the normal, non-overdriven PI balance of the two sides.
Duty-cycle modulation creates even-order harmonics and adds a dynamic feel to overdrive that changes with how hard you hit it.
Normally, this is a good thing, all old non-master Marshalls do it, some to a greater extent than others, and is an important part of the sound and feel.
However, this asymmetrical overdrive means that one side of the output stage is driven harder than the other, and sometimes this results in redplating of only one tube in a 50W or pair of tubes in a 100W amp.
This can be avoided by designing for a proper reflected load impedance of the output transformer and designing in the proper amount of sag in the power supply to keep the tubes out of the 2x dissipation curve under heavy overdrive.
Since this is difficult to change after an amp is designed, reducing the amount of duty-cycle modulation can prevent one-side redplating, at the cost of loss of dynamic harmonic feel in heavy overdrive.
Even if you don't have a one-side redplating issue, you may prefer a more symmetrical overdrive, so you can play with the value of that grid bias reference resistor and adjust to taste. Personally, I love the duty-cycle modulation and do not like a more balanced overdrive, it results in only odd-order harmonics and sounds thinner and less dynamic to my old ears.
In an amp that has a pre-phase inverter master, and/or if you don't drive the amp to output stage clipping, the effect is much more subtle. Reducing the value of the input side grid reference resistor raises the cutoff frequency of the low pass filter formed by the input coupling capacitor and that resistor, resulting in less low end and a tightening of the tone. Since the PI effectively bootstraps the circuit, the actual effective input resistance is much higher than the resistor value, typically around 4x, so you can get by with lower values without affecting the low end too much, especially considering the input coupling capacitor value is usually way higher than needed.
The RC time constant also affects the dynamic transient response, so there are a lot of interactive things going on.
Best to use your ears to determine what you prefer, unless you are trying to prevent or solve an asymmetric redplating issue.