In my previous post about a hypothetical 'reversal of time', I should have mentioned the spin echo effect.
In real experiments, first performed by E. L. Hahn, a configuration of spins evolves from an ordered into a disordered state, but subsequently the initial ordered state is recovered by application of magnetic pulses.
The spin echo effect is described in this text (sect. 11) and further discussed here.
Obviously, this effect raises several interesting questions about the foundations of statistical mechanics, e.g. the definition of entropy via coarse-graining; But as noticed by
the Aeolist "many of the terms used in the debate, beginning with the all-important definition of entropy, and including terms like ‘preparation’ and ‘reversal’ (and its cognates), are still used in so many different ways that many of the participants are speaking at cross purposes".
By the way, I did not see a discussion of the 'backreaction' of the ensemble of random spins on the magnet (and its entropy) that is used to trigger the reversal; It may or may not be important in this debate.
A somewhat related model is the swarm of n free particles moving with random but fixed velocities on a ring, as discussed by H. D. Zeh in appendix A to his book about the direction of time.