Would the bias current not be set by the potential divider formed by the feed resistor and the bias adjusting pot (in most cases) , and the two 220k resistors from the grids reference these grids to that voltage, but not draw current.
The Mullard data sheet says they should not exceed 500k.
If they were lowered in value would they not start to shunt the PI load resistors, dropping the gain of that stage.
This is all conjecture, cos i've never had problems that i thought could be caused by these resistors.
A quick look at a few EL34 circuits seem to suggest that keeping the value higher than the PI anode resistors is the thing.
Someone with more experience should come in on this one cos i dunno.