The only thing that should matter is the moment of contact. The language mentioned ("smooth," "timing," "follow-through") are descriptions of characteristics that we (as humans) need to accurately replicate the "moment of contact" for some manner of predicable result. That is, for all of us human pool players, we discuss stroke as a proxy/predictor for the "moment of contact."
If a machine could create the "moment of contact" with the same factors over and over using a 1mm, or whatever distance, "stroke" (e.g. the "spin, power, angle" noted) - it would have a repeatable and measurable result. There are no delivery/pre-delivery/post-delivery requirements. Just the force(s) at the moment of contact.
Here's a new robot that helps on this point. It does have some type of stroke, but only to accelerate the cue cue to mechanical limitations. If it could be done in a shorter distance, the result would be the same.
-td (my $0.02, probably worth less)
View attachment 878369