Well after the event whilst trying to understand the loop filter calculation better, I came across the whole newer generation of PLL chips which have a constant current source to charge the loop cap. Makes the maths much, much simpler. Even I can understand it!
Yes any ripple or noise on the control voltage of course appears as frequency or phase noise on the o/p.
Seperate supplies for the Pll/osc, control logic and o/p buffer are a good idea.
Yup the 19kc pilot tone on a Fm stereo broadcast from certain transmitters is very accurate, its also unmodulated so isnt so diffcult to process, plus with the stereo indicator you know when you have a signal and the osc isnt just free running.