I remember having the Amazon gyrate wildly at one point because its moment of inertia was ridiculously low, around three or four orders of magnitude below the retail value (and the reason for the bug was most likely that some of the values were something * 10^-16, IIRC approaching the machine epsilon for a double-precision float).
Herra just asked on IRC about the mass of the Herc, so I gave him the only value we have- PCS2 units. On a whim, I checked both the retail and mediavps values.
The retail value was 302.46. The MVP value was 3835.75. The moment of inertia is also off by, for some of the 3x3 matrix's cells, up to six orders of magnitude....
Are these values never used in FS? It seems to me that all of the mediavp values may be off by this much, probably for a long time now, and yet no one's ever reported any problems. When does the physics engine actually deal with these values where the insane differences may be noticeable? Kinetic weapons like the Morningstar, perhaps?
[EDIT]: After looking at some more MVP models, it seems that most of them got it right and are using retail values. I just happened to find two that didn't. I don't have time to check all of them now but this still may be something that should be looked at.