It's not advisable to parallel-up the VCAs. It would be better to have a transresistance output amplifier for each VCA and then sum the amplifier output voltages using conventional resistor summing or other methods.

Do the VCAs share the same input signal in your scheme or are they connected to separate input channels? Do you have the same gain control voltage for each VCA?

The VCAs use the same I/O and CV. Running in parallel just for dist/noise. I need the lowest dist/noise possible and very tight tracking between two channels, and that is why I'm looking into paralleling. The application is a precision AGC. It looks like the THATs are very good spec wise. When first building the unit I mucked up my own VCA, but gave up on the idea of using my own design. I was surprised how hard it is to design a good VCA! Theory pretty much goes out the window. I see why having an opamp on each output could be interesting, but in that case I'll be summing the four opamps THD+N and offset. I guess a simple solution is just to get the CV as clean as possible, use thermal compensation that tracks well between the two channels, and just deal with it...

If you are aiming for lowest distortion, I assume you will be using the THAT 2181A parts.

Since the inputs of the devices are virtual earths, it would not be a good idea to parallel several devices together using a common input resistor, as there is no easy way to get the input signal currents to be shared equally between the devices. However, I can't see any major problems in parallelling devices that have their own input resistors, but you would have to take care that the summed output capacitances did not give stability problems for the output amplifier. I assume that for N devices in parallel you would use an amplifier feedback resistor of 20K/N.