WSG Behavior

[reactor]

I took a few months off from working on the WSG to focus on other tasks. Getting back to it, I thought I'd made a breakthrough, but am now only more confused by the device.
The issue:
The unit works as designed by Ray, flawlessly. I ran through all of the troubleshooting steps on his site and the machine responds as it should. good job, me.
The problem stems from my need to customize the device.
I want to be able to independently switch the two voices on and off, and mix them in at varying levels. I have a two-channel crossfader schematic that works just fine. I cannot seem to get the two voices isolated on their own pre-filter on-off switches.
It seems obvious that placing the switches just before the ODD point in the schematic would do the trick. This is not the case.
Through attempts to discover the root of the trouble, I removed R2, R5, R21, and R22. Doing this should have ensured no signal reached the output. What I end up with, however, is a fully functional Voice A at about 80% of its fully connected output volume, and a similarly function Voice B, excepting that the Zany components of Voice B are around 40% volume. The filter works properly on both voices.
Why am I hearing anything at all with these four resistors disconnected?
Is there a clue in the fact that the relative volumes are not the same for all oscillators? ...and why wouldn't any of this strange behavior be evident when running through Ray's troubleshooting?

Thanks, all.

reactor.

[reactor]

I am surprised I haven't elicited any responses.
Perhaps I am the only one that has ever had this sort of problem. Nevertheless, it is my hope someone can provide a hint at why, after removing the only signal paths by which audio should reach the filter, I still have audio at the output.
Thanks to any responses, even if someone wants simply to call me a dope for not seeing the obvious (?!)

[diablojoy]

i think from memory if you simply switch a low resistance ground to the output pin of the particular osc that should work say 1k ,i can't think of where i got that from though. also try looking at super psycho lfo's so worth checking the cct on ken's CGS site for idea's he uses probably a better method though would be more difficult to implement on rays board

haha yeah it is hard to get the little mongrel to shut up
good question though with all 4 of those resistors out i would
have at first thought near silence would be all you would hear too.
however do not forget that the oscilators are oscillating between ground and the full supply voltage so perhaps that will have enough of an effect on the 741 opamp via R10 to grnd to provide a reasonable ghost.

[reactor]

I guess I'll just keep pulling resistors...

I did solder extra wires to each of the oscillator origins at the main chip, and as you suggest, dumping them to ground quiets them. So, adding a 3PST switch to each voice would allow me to switch each voice in and out with only a single switch. The problem then arises: If I can't isolate the voices pre-filter, due to some unknown signal bleed along the way, how on earth can I expect to separate the voices out to a crossfade style mixer as i earlier suggested? Theoretically, I could utilize a 3-gang rotary pot to play with the amplitude of the oscillators just over the backside of 3PSTs, but these are pricey and the whole setup just seems a bit convoluted. There ought to be a simpler way.

Thanks for a few ideas, I'll play around a bit with your suggestions tomorrow and see If I can make a little more headway.