I think I've come up with a sufficient condition for a smooth developable surface to exist between p and q. Condition: p and q can be split into a sequence of pairs of curves (p1, q1) ..., (pn, qn) such that each pk and qk * have no inflection points* turn in the same direction * have parallel starting and end ending tangents* turn the same number of timesHowever, now I see that in your construction, these splits happen at the points of inflection of the rail curves, but the tangents don't match. It seems like a conical section 'takes up the slack', as it were.Awesome pic! …