Mike's statements made me wonder if I had forgotten how the 431s behaved. My data book says Iq is 500 uA typical, and a test setup confirmed this, for both my old part and a newer one acquired two years ago. I had to look at the link he provided to see where we differ: that off state current is for when Vadj is zero volts, not just below 2.49. If you drive the adjust pin from a pot, you will find that the anode current jumps up above Vref, but as you go below Vref it stays around .5 mA and only drops slowly as the voltage decreases, not reaching zero until you go below 1V. Spice has been lying about this part. (There is a curve on the data sheet that shows this behavior .) Adding the parallel resistor can make it work, but it still doesn't appeal to me.