have you ever messed around with the impedance switch with the same load to see how different the amp sounds? It makes a huge difference in tone, the lighter the load is, the more bright/brittle, lacking bass the tone will have. Quick explanation, as I couldn't find what you were talking about in the FAQ, but the tubes are connected in parallel with each other, which means they receive the same voltage drop and the total current is controlled by the number of parallel connections. Since you remove half of the tubes, you reduced the output current by half. The two tubes that are still in the amp will be conducting the same current each regardless of how many tubes are in the amp (this is an effect of parallel circuits). What happens is that your load will not give you the same current because the voltage does not change, you have to match the load so that the current on the secondary tap matches the current you want the tubes at, and since you do not rebias and you want the current the same, that means the load has to change, it has to double. To compensate, you can use the same load if you drop the voltage in half, producing the correct current (the dissipation will now be equal on both sides).
Now unless their is something that I am forgetting offhand regarding transformers, that is correct. This also means that, if using ohms law on the situation on a load circuit, if you remove two of the tubes and keep the load unchanged, since the tubes were in parallel, the remaining tubes will double their current to compensate as the amount of current they pull will be determined by the load. This means that a 100 watt amp with 4 tubes on 16 ohms will also dissipate 100 watts with 2 tubes at 16 ohms. I can also say that is why manufactures specs for solid state amps specify the power dissipated depending on what type of load you use (ex 250w @ 16ohms, 500w @ 8 ohms, 1000w @ 4 ohms, 2000w @ 2 ohms) because SS power amps do not have a transformer, they cannot be adjusted for, so their maximum output is determined by the load.