This MR supersedes !1827 (closed)
-
So far, this only works reliably for 1p problems (without energy or mass transport). Fixing this could be arbitrarily complex and time consuming. Should we assure for 3.2 that only 1p is used or at least throw a warning otherwise?