1. Although output is balanced, it seems like the designer took an asymetrical approach as to how the phantom power is routed.
That's right and is definitely a flaw.
If I'm guessing, the FET's are getting about 10-12V & 1-2mA current after going through all the resistors from pin 2.
Actually they are operating at less tha 0.5uA, which is probably the main reason for the poor noise performance, in conjunction with the 750r source resistor.
2. The circuit seems to be made to amplify an unbalanced signal (ie the +ve side of input) as the -ve is completely grounded.
Corerct. I had not noticed it. It does not prevent the thing to operate. Seems to me like it's the same PCB as the guitar version, with its 1 Megohm input resistance. It sure works for a mic, but is not optimal.
Im uncertain if this has an undesirable effect on SNR/CMRR of the signal.
No. The LED and the way power is derived takes good care of ruining CMRR.
Before I attempt to add any additional filters to the circuit, I'd like to first attempt improvements to the SNR/Noise floor of the circuit, probably by replacing the grounded caps
The main improvement will be decreasing the source resistor to a much lower value, which would result in significantly increasing the operating current. 2SK170 wants to operate at 1mA minimum, 2-3 even better.. I would start with 68 ohms. The problem then is that you'll also need to decrease the 10k resistors.
& move around some resistors to make the circuit symetrical.
You'll need to add gate resistors to each FET. I typically use 22k, but the value is not critical.
This is what I came up with:
Pls find my suggestions.
From the start, the topology is not optimal.
It solves the two main issues of the original FetHead, but adds two other issues, poor noise performance and poor headroom.