So I think this is a really interesting design question. Mixing aside, if Quad VCA doesn’t pass its individual ins to its individual outs, but instead mutes its outputs because it’s not an “effect”, would a single VCA like VCV VCA also be expected to mute its output, rather than just becoming unity? What’s the line between effect and signal chain? A VCA with a LFO over amplitude is a tremolo and I might well want to bypass the tremolo the same way I’d want to bypass a reverb. And with fancier VCA configurations like Quad VCA, or even VCA Mixer, have both individual VCA and signal chain capability.
I think where I get tripped up is that if I think of bypassing as “removing from the signal chain” it’s not automatically clear that it means “mute” versus “remove, to the extent possible, the particular processing being done by this module so that I can hear something less processed in my output.” I was interested to note that @Eurikon’s post linked above notes that Reason offers separate Bypass and Off modes, which reflects this distinction.
Happily, the API will let module developers try different things, including implementing what I would probably do in my own modules if there were an ambiguity–a right-click configuration that let users pick a more Off-like or a more Remove-Processing-like consequence of bypass. The first would almost always mute outputs, and the second would almost always apply a “neutral” process (bypass for reverbs, unity for VCAs, unity sum for mixers, etc.)
Since bypass status will (I presume) continue to be saved in V2, there is a little bit of a patch-compatibility point here in that changing how a module handles bypassing by default may change some previously saved patches.