First of all, a cutoff of zero will let some sounds go through for a couple of reasons: 1/ there is filter tracking so the cutoff value dialed on the filter page is shifted up/down depending on which key you play on the keyboard (and depending on the patch you play on, random values, velocity, etc… check in the modulation matrix that nothing affects cutoff) ; 2/ each stage of the filter generates some noise at the output, so even with a cutoff of 10 Hz you have some background noise adding up ; 3/ a 4 poles filter attenuates with a 24dB/octave slope, not a 200dB/octave slope :)

On the SMR4 board, the VCA bleeds a bit – I have been pussyfooting over this when doing the v0.7 revision but there was no way (layout-wise) I could add the extra trimmer to keep it quiet – there is such a trimmer on the SSM2044 filter board. Note that, when the resonance is not set to 63 (ie, when the dominating sound source is the oscillators), the bleed is harder to observe since the oscillators are silenced “in software” when the slowest envelope reaches zero.

Pin 2 of IC4 is the summing node where the VCA CV is scaled / filtered / rectified. The bleed comes from the fact that the “0V” coming from the digital board is not exactly 0V, but a few mV. So you need to subtract a few mV here. Try connecting pin 2 to a 1M or 330k resistor connected to the wiper of a trimmer whose extremities are connected to 0 and -5V. The trimmer acts as a voltage divider, and through the large resistor it generates a small negative current that will suck the VCA CV down.