I read part of the "What is a Good Linear Element" paper by Professor Shewchuk at https://imr.sandia.gov/papers/imr11/shewchuk2.pdf,
and it seems like triangles with big angles (close to 180 degree) don't approximate functions well, but triangles with a small angle aren't necessarily bad.
It has to do with the fact that the gradients often get messed up with the large angles since area of the triangle can change drastically with the edge lengths not changing much.

I read part of the "What is a Good Linear Element" paper by Professor Shewchuk at https://imr.sandia.gov/papers/imr11/shewchuk2.pdf, and it seems like triangles with big angles (close to 180 degree) don't approximate functions well, but triangles with a small angle aren't necessarily bad. It has to do with the fact that the gradients often get messed up with the large angles since area of the triangle can change drastically with the edge lengths not changing much.