Why? Because the equilibrium solutions for magnetic field as a function of induced magnetization and for the force on the propellor as a function of "twist" of the rubber-band is a cubic. Notice the way those functions are going! Induced magnetization is not a FUNCTION of magnetic field (nor is "twist" a function of force) because the cubic would be "lying on its side" and we would have 3 values of induced magnetization for some values of magnetic field. Think of it as $x= y^3- 6y^2+ 9y$. The "switchback" section is between the two extrema for x, 4 and 18. In that region, the "switchback" section that connects the other two is an unstable equilibrium while the other two are stable. As you start increasing the magnetic field, you stay on the lower branch until you are past the local maximum x (in the example above, $x= 18$) and now the value jumps to the other branch. Reducing the magnetic field, you stay on the "upper" stable branch until you hit the local minimum x (in the example above, $x= 4$).