To maintain accuracy when changing Quadratic elements to Linear elements requires the element edge length to be less than half the size.

Smaller linear elements might converge where larger quadratic elements fail to converge. Itâ€™s possible that smaller quadratic elements will also converge.

Yes, the high temperature gradient is contributing to the element distortion. Smaller elements will reduce the temperature change across one element and shorter time increments will help the simulation to converge.