i'm suffering from the same problem here.
phy set rigid body rotation ID#,x#,y#,z# has definate issues, the object should not be carrying any momentum or velocities after being manually rotated to a specific orientation. even if its not a kinematic object. setting angular velocities and momentum to 0.0 on all axis seems to just have other wierd effects on the object after rotation.
can someone else please try to manually change an objects orientation before and after applying local forces to verify this problem.
you should see your object do some cool offset rotation thing, or just go nuts and start breakdancing. lol.
cheers, grant.
Ohd Chinese Ploverb say : Wise Eskimo, not eat yerrow snow.