Are we talking RANS? if yes, you must be careful of the wall functions you use for k/omega/etc, since some wall models are only valid for high Reynolds number (y+>100) and others not