It seems to be working for me using a simple side-chain test I use sometimes.
I tried the pro-channel version and the VST3 plugin version of the CA-2A compressor. Use the drop down to enable the external side-chain, add a send on the vocal track to the side-chain input (exactly what you described). Then I turn the peak reduction up 100% (just for testing). The background track gets compressed when the foreground comes in (in my case spoken dialog over a background ambient sound effects track).