Comment (by Samuel Cupp):

So, the heart of the problem comes from (I think) the function

th::regrid()

in repos/carpet/CarpetLib/src/th.cc:51. When this runs, it sets the

deltas.AT(ml).AT(rl)

object to 1.0 (possibly divided by the timereffacts), which then causes an incorrect cctk_time. This seems to be polluting some functions which call get_time. This happens without PreSync, but by miracle (design?) it gets reset to the right value before anything tries to use it. This does not occur in time with PreSync.

--
Ticket URL: https://bitbucket.org/einsteintoolkit/tickets/issues/2639/carpetlib-internal-error-with-presync-and