Hello!
It's unlikely that this will affect anyone, but I've found some incorrect behavior in the treatment of *upper-face* symmetries in ReflectionSymmetry.
I propose the attached patch. In brief, since the loop runs over the "source" points in descending order and the "destination" points in ascending order, then the index offset between the two sets should just be 2 (or 1 if one is staggering the origin), rather than twice the ghost size as in the lower-face case.
The change affects branches that are only executed upon reflection_upper_*, so most people won't see a difference. Any input?
Eloisa