Yes to everything you say. My critique of the “heaviness” you get from converting even simple SubDs to NURBS is that especially Fusion 360 more or less encourages a workflow like that… which is a nice idea in theory because SubDs are great, but in practice they result in rather unwieldy NURBS that sooner or later always destroy the parametricness of the model or cause trouble with tolerances or a million other things that can go wrong because of overdefined data. What I am trying to say is that T-Splines is hella useful but it is not the end all be all to all modelling pains like Fusion pretends it is. Because currently the deep T-Spline integration is actually the only thing Fusion has going for itself. When it comes to pure parametric modelling solidworks is quite a bit better than Fusion and when it comes to the surfacing things like Rhino are at an entirely different league. In the end designers won’t get around learning surfacing, parametric modelling and polymodelling. There is not one that can replace the other.
Which brings us full circle to the comment that learning Rhino is useless 
It is not, there are so many things that are nearly impossible or really really hard to do with parametric modelling… and vice versa.