I wonder if this means the problem is under-determined (too many degrees of freedom) and that leads to numerical instability. Certainly one set of equations is enough, as 3Blue1Brown’s Fourier transform video shows.