You bias them together and together they must stay within the acceptable range. If they don't then one or both need to be replaced.
ARC is basically saying some bias drift between the two tubes is acceptable, but when both combined drift out of the acceptable range, then one or both need to be replaced.
My question would be: how do you know which one needs to be replaced?
Mike, it's not that complicated. What Joe says is true. The tube pairs in the more recent ARC Ref amps are designed with one tube being the set tube and the other being the slave tube. And yes, as you said, tube bias may drift.
That said, unless a tube arcs, in my experience, the power tubes run out of juice at pretty much the same time. As a general matter, the KT-120s and 6550s had useful lives of 2000 hours or thereabouts. The KT-150s are supposed to run out considerably longer.
Now ... in the case of power tubes sourced from ARC, they usually bias pretty close to each other. And btw, the ARC manual says that the power tube pairs are considered in spec if the bias is between 57 and 73 mVs. IOW, the tubes are in spec if the set tube bias is set at 65 mVs (recommended value) and the slave tube bias ranges between 57 and 73 mVs. As a practical matter, my tubes spec within a 60 to 70 mV range, and most of the time it's between 63 to 68 mVs ... and they don't drift very much.
As to the question of why ARC uses this set up .... I'm not sure. I seem to recall one of the ARC tech reps telling me that ARC has a strong design bias (pun) to keep the circuit as simple and uncluttered as possible. If I had my druthers, as a consumer, I would rather have individual tube bias adjustment capability.
Bruce
