Baking environment values into the actual deployment artifact also doesn’t pass the 12 factor smell test for me, but I may be revealing my lack of knowledge of their architecture / process here, maybe this is how they get repeatability.
I don't think this is a good idea. Please convince me otherwise.