In any case, if by "Liouville's theorem holds", you just mean that you can take the evolution equation with the Poisson bracket instead of the Moyal bracket (because all the higher order terms in the Moyal bracket vanish in the case of the HO Hamiltonian), that's one of the many ways in which the harmonic oscillator is "semiclassically exact", i.e. you can often get the correct answers by just taking everything to first order in $\hbar$ or something like that