The following discussion from !161 (merged) should be addressed:

• @sospinar started a discussion: (+2 comments)

  The normal vector is positive to the outside of the cell. That means that this the negative value of \frac{\partial u}{\partial \vec{n}}. Of course this is doing the same as before, but I find it more clear if it has the correct sign (as the DG local operator).

The resulting change in 5e15de0e should also have inverted the checks for upwinding. It did not, which is why the upwinding now effectively is a downwinding in the FV solver. The ODE tests did not notice that change, which becomes mostly apparent in transient states.