Anyhow, suppose we have magnetic field in the z-direction which is +B when x>0 and -B when x<0. In that case, a "Landau level" n in the right halfspace, with spin parameter -, automagically transitions to a Landau level n with spin parameter +, in the left halfspace, whenever cross the x=0 plane. I.e. the spin reverses with respect to the B-field, but aside from that the Landau states stay the same whenever cross the plane. This corresponds to an energy step up (or down) by 1 level. Nonrelativistically this step size is constant hbar*e*B / (m*c) but relativistically it is nonconstant i.e. depends upon n and decreases for n large, ultimately to 0. We regard the "Dirac sea" in this scenario as consisting entirely of Landau levels. In that case we have simply a potential step, from the view of the two Dirac seas. If because B is large, the size of this step exceeds 2*m*c^2, then it seems to me we ought to get pair production since a negative energy state in sea#1, can have positive energy in sea #2. Furthermore, the rate of particle production should be calculable since there are only a finite set of Landau levels for which the step size exceeds the threshold; we can sum over all of them therefore.
--And... apparently this scenario never can create a pair, because a negative Landau energy level never can become a positive one in this way. You can gain energy from moving to a region with different magnetic field; thus you can feel effective force in a magnetic field gradient-- even huge force; but it cannot promote negative energy states in the "Dirac (Landau) sea" to positive energy states. So... it is looking like I am coming round to the boring view: magnetic fields cannot create pairs.