This is problem has been more or less completely open for over 150 years. There have been some good results for $r =2$ and sporadic constructions for $r \geq 3$. Evidently Peter Keevash has cracked the problem wide open by solving the case of general $q$ and $r$. I'm not sure that anyone saw this sort of generalized construction coming. What's more, it seems that there is a new probabilistic construction technique at the heart of the proof. Incredible!