The article is poor. I think the research is another take on how finite reversible systems end up cycling. Their entropy has to eventually decrease because as

The article is poor. I think the research is another take on how finite reversible systems end up cycling. Their entropy has to eventually decrease because as a path through the phase space runs out of places to go it must return to the initial state to avoid getting stuck (i.e. violating reversibility). I'm guessing that the interesting new thing here is that the researchers showed systems following gravity have this entropy-goes-up-and-down property, despite being an infinite continuous phase space where the bits could just get more and more dispersed? Of course in the actual universe there's other forces at work and they need to be accounted for. You can't just say "gravity creates low entropy times therefore we have explained the low entropy", because things like the accelerating expansion of space have to be accounted for. You also need to account for how incredibly low entropy was near the big bang: it wasn't just a single galaxy's worth of stuff converging, which would have been sufficient and far far more likely, it was billions of times more. Sean Carroll frequently talks about this issue in has talks and on his blog. Hopefully I didn't get the details completely wrong.