The fluctuation theorem work is not about defying the 2nd law. It is about deriving the 2nd law as a classical limit of a deeper statistical mechanics law. As their name indicates, they are about fluctuations, often fluctuations about an equilibrium. The question they address is, statistically, how long would you expect fluctuations away from equilibrium to last. Even these small systems do not decrease entropy for long, nor predictably.