I think the linear least squares is like a shear, whereas the eigenvector is a rotation.
Shear with one component always in the y direction
Shear with one component always in the y direction