I may be a hack, but I switched from using 4-gang pots to shunting stereo pots because I think they work better.
A big part of the reason why I would build a balanced circuit is to benefit from its CMRR (and even order distortion cancellation by proxy). I measured the matching between gangs on a 20k RK27; it's better than the specified 20% tolerange, but it's not great. Here are some measurements at arbitrary rotations.
Gang A
Gang B
Matching
51.4
51.6
0.4%
255.7
252.8
-1.1%
623
624
0.2%
1299
1300
0.1%
3042
3089
1.5%
4230
4368
3.2%
8470
9080
6.7%
14680
15420
4.8%
19100
19930
4.2%
19900
20380
2.4%
The matching is directly proprotional to the CMRR. So I took a measurement of CMRR using the following circuit.
The idea is that this simulates the balanced input of an amplifier and will tell us how balanced things actually are.
Not very, it seems. Each trace is a different rotation angle on the pot, with the -115dB one being with the signal fully shunted. (There is some resitual resistance with the pot at 0.) -35dB of CMRR will significantly degrade an amp's performance.
I repeated the measurements with a shunting pot configuration, using Yageo 1% 2k resistors and the same 20k pot.
Although the AP input capacitance degrades CMRR at high frequencies, even the worst 1kHz measurement is about 12dB better than the best measurement with the series pot. The low frequency degradation is most likely caused by the AP's 1/f noise-- we're measuring microvolts at this point.
If I wanted to further improve performance, I'd use resistors with better matching. The Vishay ORN arrays are available in 0.01% and would probably be my pick. LT5400 isn't bad either.
https://www.mouser.ca/ProductDetail/71-ORNTA1001ZUF
I don't know that this is necessarily relevant to the design in the first post-- looks like it doesn't use a pot at all-- but the circuit has its place.