Norman, I deleted the 200-vertex objects leaving the 40-vertex ones - which compile without error. Except when the object fills the entire screen (in FSX), my eye cannot discern the segmented edges. And even then, I've got to look very critically.
200 vertices in a larger object will be OK. But, for an 8' circle, the resultant triangles have at least one side only 1.5 inches long. The other two sides could be up to 8' long with a subtended angle of only a fraction of a degree - in some cases only a tiny fraction of a degree. (Triangulation takes place between vertices, not from the center). With these length/angles, even the curvature of the earth plays a role.
To avoid other issues, distances/differences of less than 1mm are ignored. From my tests on your airport, I know it is the first triangle that causes the grief. What this tells me is that the width of that triangle (due to the subtended angle) is less than 1 mm., i.e., effectively 0. That's a straight line, not a triangle, so the process fails.
I'll leave it to you to determine just how many vertices you can use before the ADE_GP compiler "gives up", but 200 is definitely "on the wrong side" for an 8' circle.
Don