Comment (by Roland Haas):

Hmm, I cannot say I have a good suggestion how to proceed, I am afraid. If this is just around the time that the stars come into contact then it could be a con2prim failure which is very hard to debug, in particular since I do not think that the cold tabulated EOS has been heavily used for binary simulations.

--
Ticket URL: https://bitbucket.org/einsteintoolkit/tickets/issues/2621/keep-running-into-nan-need-help