Comment (by Roland Haas):

The remaining issue does not seem to have been due to unitialized variables. Namely the effect in my BBH test run seems to be an exponential growth of oscillations in the conformal factor in all the physical boundary points (with or without the z-symmetry). At least it does not trigger any of the unitialized data checks in Llama itself that already fills the multipatch which exists in Interpolate2 all the time:

738:     // interpolation source
739:     if (verbose) CCTK_INFO ("Poisoning inter-patch boundaries");
740:
741:     {
742:       list<scatter_setup_t>::const_iterator scatter_setup_it =
743:         interp2_setup.scatter_setups.begin();
744:       BEGIN_LOCAL_MAP_LOOP (cctkGH, CCTK_GF) {
745:         BEGIN_LOCAL_COMPONENT_LOOP (cctkGH, CCTK_GF) --

Ticket URL: https://bitbucket.org/einsteintoolkit/tickets/issues/2030/multi-block-boundaries-leave-uninitialized