As I read the spec, it is definitely OK to specify one input that controls the polyphony, and then all the other inputs replicate as needed to match. That seems to be how most poly modules work.
That is why I was excited about how the V2 VCV VCA works - it indeed implements “The Input with the highest number of polyphonic channels defines the overall polyphonic channels of the module”. I don’t think that feature is very common.