I have a double sided 2D grid of squares with four corner vertices and four edges. I find the points within the radius and build up two hemispheres with polygons within the radius, XY and an assumed Z coordinate.

This is all well and good, but I should probably go back to dividing a sinusoid into diamonds and triangle pairs and then group them into subdivided quad/triangle pair corners. What I was doing with my comparable area rows of quads within a sinusoid is also not a bad idea, for that survey setting I was working on all those years ago...