That's a good question i thought i made a mistake...After some relfexion, I would say : if the ang. speed. is constant, then the body does not need to be rigid...in fact i think, in a real case, the body will modify itself during ang accel. at the beginning, and then reach an equilib. but this equib. final state at constant ang. speed is different than the body before you started to put it in rotation (maybe not for jelly, but for soft chocolate)...so the answer is : if you want same constant ang. speed (hence not the instantaneous one in an accel. case), it does not need to be rigid...but if you have an ang. accel, it needs...i think it's more correct, but there still lack some calculation, for example : does the ang. accel. to be everywhere the same need maybe not to be completely rigid, but some kind of special deformation are allowed...