Optimizing a p3 solution for area turns out to be pretty addictive. Early on, I had a rotation-based solution using big multiarms which was fairly compact at around 70 area, but a lot of space was wasted from length-3 swings. A translation-based design allows the glyphs to be placed closer together. The challenge is then trying to minimize the number of tracks required, and sharing track loops if possible. There is also the option of doing it freespace-I style with a single long trackloop, but I don't see it being a better approach.