Ah. I'm a goose.
The codebase still has references to the zxing calls, but swapped out using it without removing the old code, which led me down the wrong path.
Before it changed, it was using AVCaptureVideoDataOutput to process the frames itself.
Then it changed to using AVCaptureMetadataOutput instead, delegating the image processing to AV Foundation instead of zxing.
Looks like to add codabar support, I may need to reverse this, since AV Foundation doesn't do codabar.
Not ideal, since I imagine the original devs had a good reason for swapping it out in the first place...
The git commit links to a ticket with the whys and wherefores.