The angelfire site is interesting reading, and some of the subject would be am worthwhile modelling task on Wayne Johnsons CAMRAD II software by NASA and the USMC. The lateral roll rates are normally not that high, but it is true that the anti roll input will increase the potential for recirculation from the instantaneous increase in the outflow velocity before the roll rate starts to reduce. As a percentage additive of the total flow conditions they are relatively small however, so are aggravating but I would doubt that they cause the problem if a VRS condition is being approached already.