Look at valves with a very different g2 voltage rating fro the plate. 6146 is one. Most of the TV sweep tubes for another. If it were desired to operate these types with a distributed load, or Ultra-Linear and feed g2 at the same potential as the anode, the maximum voltage would be severely limited.

If we instead wrapped a seperate center tapped winding long with the main anode coil, we'd get the same sort of plate feedback to the g2 as with U-L but we could run g2 at a differnet DC voltage.

Or...look at the Quad II OPTx, it has the third CT winding connected reverse phase ( compared to the previous example ) and the cathodes are connected to its ends. This has two effects. First is negative feedback delivered as the cathode follows the input signal. The other is an increased load ( nominally a-a ). The valve in the middle does not recognize how much of the load is on its cathode and which is on the anode, but the overall length is increased by ~16%. The load impedance is increased by 1.16-squared times the original anode load. Also Quad discovered a reduction in third harmonic beyond that offered by the level of NFB provided by the cathode laod. The cathode winding includes g2 current and this seemingly small contribution has important effects since PP only deals with elinimation of even harmonics.

There's no doubt something I left out as obvious to me...
regards,
Douglas

All this is overshadowed by voltage ratings between the anode winding and the cathode winding is hard to deal with. McIntosh has three windings, wound tri-filar and those old black amps seem to survive well enough.