The idea is that the pressure solver is more stable and can still handle high subsonic velocities accurately. nOn the other hand, the density based solver becomes unstable at low velocities and the solution accuracy is doubtful. This because of dealing with lots of small terms that approach 0.n