Do some sort of debugging of which sample points are usually triggering the max error

Is it always the center point? Is it ever the sides? Should we be sampling other points on the interior of a triangle?

Or is there some way to visualize such a scalar field? Like visualize the magnitude of error within a triangle?