The test ReflectionSymmetry/KerrSchild-rotating-180-EE.par depends on initial data from KerrSchild and NoExcision (which includes solving an elliptic equation). This means that problems in either of those are caught by the symmetry thorn test. The symmetry thorn test should rely on as few other thorns as possible, and those should be as simple as possible. We need variables with nontrivial parities etc, but we could implement those in ReflectionSymmetry itself, or use McLachlan with simple initial data. I propose that these tests are decoupled from the complicated thorns.