I noticed a very strange thing about this unit that I want to pass on. In final tests I saw that on both channels, turning down the output level increased the amount of compression. This was not good, and initially made no sense at all!
Long story short, it turns out that the output amplifier in this unit is not a fully buffered low impedance driver, as one expects in solid state gear. The load on the output of the amp reflects back to its input. The output control is a constant impedance attenuator between the output amp and the output terminals, but it's only constant impedance if the output is terminated with 600 ohms. Which mine wasn't.
With no output termination the amp sees 600 ohms with the output pot all the way down, but with the pot full up it sees a much higher impedance. Therefore, when the pot is full up the output level of the output amp increases, as does the input
to the output amp (strange, but true). The input of the output amp is connected to the input of the sidechain circuit. So turning the output pot down increased level to the sidechain circuit, causing more limiting.
The solution was simply a 649 ohm resistor across the output terminals. I use 649 because, in parallel with a bridging 10K load, the total is close to 600 ohms. I soldered it to the inside of the unit so that it wouldn't inadvertently be removed. This totally resolved the problem. Very strange, I must say! In 35 years of tech work, I had never seen a limiter do this before.
I'm filing this one under "design error".