Here is an argument based on spacetime symmetry and the fact that U(1) theory is abelian (therefore photons don't interact with each other) to show that the probability is zero.
Suppose we start from a uniform magnetic field in the $z$ direction. Since there is no preferred direction in the $xy$-plane, the final state photon must be emitted along the $z$ direction. The final state magnetic field must be uniform along the $z$ direction as well, due to the same argument.
Therefore if the probability is non-zero, a uniform magnetic field will continue to decay spontaneously until its strength approaches zero (as we approach $T=\infty$), and produce a collection of photons. Therefore the end state at $T=\infty$ will simply be a collection of photons, ie. $a^\dagger(\omega_1)\cdots a^\dagger(\omega_n)|0\rangle$.
Now QED is time symmetric. Therefore we can start from the final collection of photons and evolve backwards in time. However since in U(1) theories gauge particles do not interact with each other, we simply have a collection of free particles --- the interaction Hamiltonian is zero, and the initial state must be a collection of free particles as well. QED.
EDIT: Before anybody starts babbling about "non-linear" optics and "four-photon scattering", let me point out that the current experimentally accessible field intensity is well below the Schwinger limit , at which point we expect new physics to come out. Our current experimentally verified theory contains only linear photon interactions. As theorists tend to get carried away with formalism we should always remind ourselves that physics is ultimately an experimental science, and any theoretical formalism is ultimately an effective theory.