Some VCAs do leak. But most decent-quality modern ones (like what Erica uses) don't...at least, not without a reason. Another thing worth checking is whether there's an accidental low-level offset voltage getting into the control side, as this will definitely cause audio leakage from the VCAs not being able to fully close down. It doesn't take MUCH...just a tiny fraction of a volt.
Another possibility could be voltage from an outside source. Case in point: many years back, I was working with our local community radio station in several capacities. One of these was maintenance. One evening after close-down, I and another engineer were trying to figure out why the station's Harrison AIR7 would suddenly switch channels off or on without warning. It was definitely a primo radio console in its day (hell, I've used Harrison desks as far back as my first year of undergrad), so the behavior was downright weird. But it's also worth noting that, when it was installed, they also put in the remote-start lines...but then never connected them to the turntables, cart decks, etc. So they were just hanging there...cables to nowhere, only a few feet long.
At the same time, we also had an FM station dead-smack in the center of town, pumping out 10 kW. I knew already that it had an annoying tendency to creep into audio lines both at our station and at other broadcast facilities in the area. So, how could that be an issue here?
Well, it so happened that, after a bit of antenna length calculation on my part, the disconnected remote-start lines...which, of course, also connected to each channel's ON/OFF logic...were cut so that they were all around a quarter-wave's length at 92.5 MHz, which was the other station's operating frequency. After I cut all of those (as they weren't in use anyway), the "ghost" of the AIR7 was successfully exorcised.
So what the hell was going on? Simple. There was enough induced voltage on each wire that it would send voltage back to the board's ON/OFF logic for each strip. And when there was a particularly loud signal on the one station, it would "hi-state" the strip's logic...which in this case, was being interpreted as an OFF signal.
So you might check around for sources of RF in addition to the typical synth bugaboos. After all, it doesn't take very much voltage to open a VCA a tiny bit...even less than I had to chase down in the story above.