I finally solved the problem with my KSB2 board. It turned out to be a poorly soldered joint on one of the transformers that terminate the filter. I found it by writing down everything I knew about the problem and using that to systematically exclude different elements in the system.
The final clue was that receive audio was poorer through the KSB2 filter than through the variable filter on the RF board. Poorer, but not absent. That pointed me back to the filter.
I was ready to rip the filter out and rebuild it from scratch, but decided to have one more look at all the joints using an X3 magnifier. I could just see that one of the leads from T1 primary had an annulus of solder and that I could see the copper end of the wire. Sussed: I obviously hadn’t fully stripped the insulation when I made the transformer. A continuity tester confirmed the problem: high, but not infinite, resistance.
It was a 5 second job to re-solder the joint: leaving the iron on long enough for the remaining insulation to bubble off. Just for completeness, I re-did all the other solder joints as well.
The result is that the board now works fine. I’m not completely happy with the performance, or of the filter alignment, but that’s for the next post.