In Geometry expressions, if you (for example) create a segment AB and a segment CD and then constrain them to be perpendicular, the drawing responds asymmetrically to dragging A,B,C,D. (There are two modes, I forget exactly what they are and which of the vertices respond in which of the ways, but it's a quick experiment to redo.)
Here's a rather more extreme example. If you set up the following, with distances DE and AC fixed, then you can no longer drag point B at all, but you can slide F back and forth along DE, in which case B does in fact move along a half-ellipse. In dyna3, we would definitely like to be able to drag B around the full ellipse or move F back and forth along DE, which would presumably leave you stuck on your "current" half of the ellipse. Ideally you should then be able to define or at least display the full locus of B all at once -- this should be a, or even the, natural way to draw an ellipse in dyna3 (I mean except for the fact that at some point ellipsoids and hence ellipses as the intersection of a plane and ellipsoid could quite likely become primitives in dyna3).
Comments on how motions should work to be "symmetric" would also be welcome here. For example, suppose you have constrained A,B,C to line on line l with AB = BC. How should each of A,B,C move when you drag any one of A,B,C? The answer should be the same for A,C, but could differ for B. (And of course, all three should derive from some general mechanism of how to perturb constructions.)