Comment (by Roland Haas):

I had a chat with @pmoesta who is using GRHydro for MHD simulations. Philipp had a look at your parfile and pointed out two issues that I had missed:

1. you need to set GRHydro::transport_constraints = yes otherwise the default is a free B field evolution which is unstable and quite likely very much untested.
2. you should set HydroBase::initial_Bvec = "zero" rather than None (unless some code of your own actually sets Bvec to zero). Otherwise Bvec is not initialized and would retain the poison value that Carpet sets it two when allocating memory for it (or even worse it would be left unitalized which could be any random value if poisoning is not used).

--
Ticket URL: https://bitbucket.org/einsteintoolkit/tickets/issues/2272/grhydro-c2p-inconsistency-with-and-without