I’ll take this one:
Which specific capabilities do you need?
Advance sub-d for accurate design. Seek to develop sub-d workflow with tools to enable generation of forms with specific region parameter numerical definition.
While such can be done presently with existing sub-d tools (convert, match, snap to surface, loft) value will be added by any advance to the simplicity, speed, and effectiveness at numerically defining and controlling specific localized regions of a larger sub-d, especially if interactive with curve detail dimensional feedback/definition, rather than simple push/pull distance, etc.
Think network surf. One can define the radius of a curve to x, and the resulting surface will interpolate something very close at the curve region, if defining curves are adequate to do so.
Consider potential techniques to enable Rhino precision nurbs design when defining regions within a complex sub-d.
What specific problems do they solve for you?
As everyone knows…the problem with nurbs is the time it takes to define complex polysurfaces, and then the exorbitant time to iterate, edit, and/or make tweaks.
The problem with sub-d is efficient definition of design accurate forms - accurate defined as “this area needs to be x dim, radius, etc.”
Merge key elements of both, as best as humanly possible, to advance the art.