Should keep in mind that the meter combined with a PPIMV at various settings can be misleading when reading the negative bias voltage as opposed to bias current. Meters have a very high internal impedance when measuring DC, usually many Megs, but cheaper meters may have much lower internal resistance, and regardless it can still throw of measurements of - bias at pin 5s of the power tubes.
Example say a you have a cheap meter, specs say only 1Meg input impedance for DCV measurements. This means the meter itself looks like a 1Meg resistor and forms a voltage divider with the PPIMV pot (or just the 220Ks if no PPIMV is used) when measuring for negative DC at power tube pin 5s.
Ignoring grid stoppers, this means talking a measurement at pin 5 with PPIMV maxed (or a stock no PPIMV set up), you now have a 220K/1Meg voltage divider that otherwise wouldn't be there in normal operation (if meter was not there talking a measurement). With PPIMV off, the voltage divider disappears as the power tube grids and 1 Meg of the meter are now directly shorted to the -bias source.
Ignoring the 1Meg in parallel to the bias pot and resistor here, with PPIMV off/set at zero, say you measure -45V with this meter at pin 5s. Assuming there is no grid leakage from the power tubes, you'd then expect to see attenuation of that voltage reading when the PPIMV is then cranked to max. With this meter, attenuator/voltage divider is 220K/1000K now approx, so you'd expect the voltage reading to be reduced by 220k/1220K=18%, so would read about -36.9V now on pin 5s. So here the meter itself being in the circuit is changing your readings. If the meter was not there, the voltage divider would not be there and thus there would still be the original -45V on the grids PPIMV whether cranked or set to 0.
A better meter may have say a 10Meg input impedance. So say the same deal, measure -45V with PPIMV off, now have a 220K/10,220K voltage divider, so would expect to see the reading drop from -45V by 22K/10220K=2.15%, so about only a volt drop now to -44V. Again if the meter was not there, as would be the case when the amp is operating as normal, this drop would not be/isn't there either, voltage really is still at-45V.
All that said grid leakage can/does happen too, can follow Toner's link above to read up on that I think, but the main point is that the meter itself and quality of it can be a very misleading thing regarding pin 5 readings with different PPIMV settings. Meter specs effect how much the PPIMV setting seems to effect the - bias voltage reading. Better meters will read less change, cheaper meters more change, and that change you see well may not really there in normal operation with the meter out of the circuit - at least most of it as there could be/is some grid leakage in play.
Since grid leakage can be/is still a factor, that's the main reason you should bias with PPIMV maxed IMO, but the above is more for those of you with huge differences in - bias voltage readings like Toner had and who this response is mainly for (should have elaborated on this meter issue first time you brought up your large - voltage change). That - voltage change may well not be anything to worry about, mostly not even really there, just your meter throwing readings off. Best test is to keep an eye on the bias reading across the 1 ohmer, or whatever method you use, to check how the bias
current itself changes through the PPIMV sweep. That eliminates the meter error in just checking pin 5s, and you'll know if your bias current does or doesn't really change through the PPIMV sweep, how much grid leakage plays a part.
Hope that made sense.