I have used X-Plane as part of my Masters degree and subsequent consulting work, which involved an investigation of Unmanned Aerial Vehicle flying qualities and performance. In general, X-Plane gave very good lateral/directional stability results for rudder doublets (directional damping response), spiral stability and roll rate time constants. Basically I validated the simulation with flight test results for a small fixed pitch/fixed gear light aircraft which was instrumented for this purpose. X-Plane can provide corresponding "flight test" data as an output, which can be plotted, analysed and compared, which is feature that I understand that MSFS can not. I found that aside from being a fun flight simulator, it is a useful engineering tool, when applied in the right context.
Getting back to original post, I suppose that to get any real feedback on how closely MSFS models the 737 or other aircraft, you have to set up a series of controlled tests specifically to gain that data. This is lot harder than it seems.
The question is, how realistic do you have to be?
Regards
Woxman
PS. My Thesis compared over twenty flying quality and performance measures (in addition to lateral/directional stability) as part of the X-Plane validtion process. The results were surprisingly accurate...