Also remember the Mach number effects on the static stability of the airplane. This is what Bill Gunston was talking about. As the airplane goes past Mach 1 it can gain another 20 or 30 percent or so stability (at least on the airplanes I have worked on) Equivalent to moving the CG that much subsonically. So if you have a airplane that is going to spend a lot of time supersonically and want to reduce the trim drag then you adjust the CG to make it just a little stable supersonically. Then when you look at the subsonic case you are a lot unstable. With modern autopilots you can handle the instability.

The Harrier has the anhedral in the wings (well a big driver no doubt) to get the Clbeta (rolling moment due to sideslip) term to a lower level. A shoulder wing airplane like the Harrier gets a lot of yaw into roll coupling which may or maynot desirable. It depends on what the designers wanted. For example it is great in a rudder-elevator airplane but not in a pattern ship.

A certain level of Clbeta is desireable and you generally try to work out how much before flight based on similator work and pilots comments. "I can't fly this piece of _______" gets a lot of attention.

Of course another thing that gets into the compromise is the length of the wing tip wheels and positions of air to ground stores on the wings. All airplanes seem to be compromises that result as designers work and bicker to get their way.