For a long time, I thought the effect of the "color adjust" variable capacitor on the XT motherboard was to change the duty cycle of the 14MHz clock signal going into the CGA card. This would in turn change the hue of the green and magenta chroma colours (and all the artifact colours) while leaving the yellow, blue, red and cyan artifact colours alone.
However, having now tried it out, this does not seem to be the case at all:
The hues of all the chroma colours are changed. I don't really understand why, since the yellow and color burst signals are the same, so there's no way to introduce a phase shift between them (and hence no way to change the hue of that colour).
Looking at information about crystal oscillator circuits, a variable capacitor like that is usually included in order to fine-tune the frequency of the circuit. Quartz crystal frequency sources have a very small damping factor (high Q value) which means there's a very narrow range of frequencies in which they like to oscillate, but the external circuitry can influence the output frequency a little bit (as we've seen before).
But what we're seeing on the screen there doesn't look like the effect of changing the frequency either. With a frequency that isn't quite right, I'd expect to see hue changes on the right-hand side of the screen but not the left.
That makes me think that the TV must be using both the phase and frequency of the detected color burst for its reference. I'm not sure why it would do that - it seems like it would be more trouble than it's worth. Maybe it was too difficult to make an oscillator that retained the right frequency with sufficient accuracy that the right hand side of the screen never exhibited a hue drift with respect to the left over the entire lifetime of the TV and in all the conditions in which it might operate.
And that still doesn't explain why the hue of the yellow bar changes (the effect is the same even if I put it at on the left-hand side of the screen). Perhaps it's to do with the TV expecting 227.5 color carrier cycles per scanline and receiving 228.
I guess I can find out for sure whether the effect of the capacitor is frequency or not by using a second 3.57MHz oscillator connected to the TV and observing how the beat pattern changes as the capacitor is turned.