Strictly speaking, the expansion link needs to be perpendicular to the centre line of the motion at front and back dead centre. As you say at front and back dead centre, if the gear is properly set up, in theory the die block can be moved from top to bottom and vice versa without moving the valve. Angularity applies to the eccentric rod as well as the connecting rod. As the NYMR diagrams basically show, the angles of swing of the expansion link fore and aft have to be equal which is usually achieved through the so-called backset by locating the drive pin to the rear of the expansion link fulcrum. As an aside, it would be interesting to know what the valve diagrams of the Southern 2-6-0s look like as they appear to have no backset. The NYMR diagrams are copied from Yoder and Wharen (fig .170 p.158). This whole chapter of Yoder and Wharen (Locomotive Valves and Valve Gears, dating back to 1917) is on distortions of Walschaerts and measures that can be taken to mitigate them including the method of die block suspension. I suspect that the problem the GWR sought to address is derived from angularity, but the purpose of the cranking of the King rocker remains a mystery to me.