Theoretically waveform relaxation should reduce the error for a scenario with subcycling to the order of magnitude that we observe without subcycling. Is this true? We should check this!

Cases to check

for cases see fenics-fenics/README.md

Machine precision can be reached

• WR11
• WR12 (Neumann solver subcycles; temperature at interface is subcycled, flux is interpolated)
• WR21 (Dirichlet solver subcycles; flux at interface is subcycled, temperature is interpolated)
• WR22 (Both subcycle; temperature and flux at interface are subcycled)

Identical results for Quasi-Newton (compare precice-*-iterations.log of scenario with and without WR)

• WR11
• WR22 (can we expect identical results here? See Reduce error with waveform relaxation + subcycling #21 (comment))