> If you were to simply open that one switch, the input impedance would increase. Apparently, the designer felt it was important in this case to maintain the same input impedance--so when the upper switch is open, the lower switch closes and grounds one end of R2 and the input impedance (R1 || R2) remains the same.
I think it is even stupider than that. Without switch CV1, both ends of R2 would follow input voltage. Switch CV2, now "open", has ground on one side, raw signal on the other side. And the CD4016 is a pretty crude part. If run from +/-5V (or +5/0!) rails, and +/-10V signals on R2, it will break-down and load the input and leak into the summing node. By adding switch CV1, the right end of R2 is always "at ground", so switch CV2 is never stressed.
You could feed CD4016 +/-9V, but signals can go higher. Also the CD4016 has no logic level translation, so high rail voltage means high logic levels.
A similar 2-switch plan is used in some channel muting schemes. An open series switch mutes the feed, but the open switch won't stay open with large signal swing. A shunt switch keeps the opened connection close to ground, avoiding leakage and breakdown in the series switch. Yeah it needs an extra switch, but a CD4016 has four and only costs $0.16: 4 cents a switch.
There is some reason to use the original flavor 4016 instead of the new-improved 4066. Generally these are exact replacements for each other. 4066 has lower resistance. But they do that by "cancelling" Rds(on) variations. So the resistance of 4066, while lower, is lumpier as signal voltage swings. The 4016 resistance is higher, but varies more smoothly, less annoying distortion. (And at one time, 4016 was cheaper; today both are nearly the same.)
