If all else fails, I could probably provide a simple, randomized mesh for testing. It would just be a flat table cloth or something like that. I've got some Java code for computing Delaunay triangulations flying around from my teaching days.

-- I'm not mad at you, just Westphalian.