I thought that wing dihedral provided some stability in roll. The wing which - as a result of a disturbance around the roll axis - becomes more horizontal is going to produce more lift, and thus counter the roll.
A wing that's "more horizontal" doesn't produce
more lift. It just produces its lift in a direction closer to the vertical. But if you take both wings together, the direction of the resultant simply rotates with the aircraft. There's no net moment about the roll axis.
You can also envisage angular displacements that are not strictly about the roll axis but about an axis displaced from it by the angle of attack. Such displacements do result in a rolling moment, but only by introducing slip. The fundamental issue is that an airframe (with side to side symmetry) has no way of knowing which way up it is, only of its slip or pitch angle, and how much it is yawing, rolling or pitching.