After studying Rev 3 for quit awhile I found that the main AC current leakage path was the drain-gate capacitance of Q1. The drain-gate capacitance of Q1 would couple into C2 and then go right out the bottom of the CCS.

With the adition of Q2 and C5 the leakage is now "captured" and routed back into the signal current at the source of Q1. The source of Q1 is the summing node of the CCS. The bias for Q2 is developed by adding a resistor in series with R4. Q2 and the pentode does not need a squeeky clean reference.

The addition of Q2 and C5 increased the performance from the 200M ohms and 3pf shunt capacitance of the Rev 3 circuit to 2G ohms and .15 pf shunt capacitance.

Listening tests confirm that the extra isolation is quite noticable.