I'm confident they could come up with a filler eye animation algorithm that was convincing enough to pass muster for short periods of time. Even if hand coding something didn't quite work out, they certainly have tons of eye tracking data internally they could use to train a small model, or optimize parameters.