It took a couple of days of reading and bench time but I finally made my 4066 bypass pop free. I thought I would post this for anyone else who might be having problems.
Things I figured out.
1) If you try to measure a bias voltage that is fed with a high value resistor your DMM will probably load the circuit give you an incorrect reading. If I have a 1 M resistor to the 1/2 V+ point and try to measure that with a DMM with a input impedance of 10 M, the loading is significant in this case. The voltages I was looking at were in the millivolt range, so it makes more sense to measure the voltage across the switching element, not with respect to ground.
2) The inputs and outputs need to be biased to 1/2 V+
3) The resistors used to bias the inputs and outputs can't be too large. I had 1 M resistors in there and the switch popped, lowering them to 100k got rid of the pop. Something to do with internal capacitance of the IC, explained by RG elsewhere (something about charge injection). I have built a 4053 bypass with 1 M bias resistors and it didn't pop, not sure it you need to check this on a case to case basis. (I just noticed on page 103 of "The Stompbox Cookbook" there is circuit snipped of a 4066 switch with 10k bias resistors)
4) After that I still had a slight pop, and I measured 22mV across a switch element. I finally figured out that 220 nA was flowing through a 100k resistor to product that voltage. It was the bias current for the NE5532 that I had put after the 4066. I replaced that with a TL072 and the noise is gone.
Nothing original here, I must have read all these points ten times elsewhere before they sunk in.