Segregated or coupled solver?

Nori

I have a 10" X 10" chamber with subsonic hot gas at 2400F and a pressure of 650psi entering the domain. After the chamber the hot gas goes through a 1" hole that is 3/16" thick. Here the hot gas increases to supersonic velocity. From here it exhausts into a 10" X 50" chamber at atmospheric air and 80F. I have read in Fluent that the segregated solver can handle supersonic flows to a certain extend, but the coupled is better at supersonic flows. I have started my case using the segregated solver and using FMG initialization using default under-relaxiation. After about a couple thousand iterations the case divergeres. So I decreased the under-relaxation for all and tried again. Again, after thousands of iterations the case diverges. Now, I am using coupled solver and it seems may case is much more stable with residuals decreasing. Can someone explain why segregated isn't working? Everyone I have talked too says that I should be using the segregated solver since my inlet flow is subsonic.