The angle distortions at the second and third vertices in each triangle seem to have been mixed up in the previous commit. I've checked, for a few triangles, that the output of this commit is consistent with the law of sines.
I've confirmed that this commit's output matches the previous commit's up to ordering.