No, thank you for the thought, but I’m only interested in a tangent to the circle that is the edge of the cylinder. And as I said above, replacing the edge in this with a circle doesn’t solve it.
And the corner of the box should be nicely in the plane of the edge/circle because the box was derived from the cplane bounding box of the cylinder. But your post got me thinking, “What if it isn’t?” And then I remembered that there is a recent thread about the bounding box being inaccurate in R8 (ERROR Bounding box is off in one direction - doesn’t seem to have been picked up by McNeel yet, @pascal ). So I checked and the box is indeed inaccurate (and a Near snap is obviously not an acceptable snap here).
That made me wonder how close to the plane of the circle a snap needs to be to work. In this case a snap up to 0.004mm from the plane works, 0.005mm doesn’t (I haven’t tried intermediate points). That is in the context of a model tolerance of 0.001mm.
@Pascal: Given the tangent is held so rigidly to the plane of the circle, I suggest there is a case for allowing its second point to snap to a curve’s Near snap point that lies in that plane, as connecting to that curve is the likely intent of the user.
In the meantime, the solution is to create the plane and put a point on it where it intersects the curve. You can then snap the second point of the tangent to that point.
Thanks for jump starting my brain @AlW!