Nvidia suggests CUDA-gdb for this purposeI do not know if this would work with this kind of computation but I would suggest you try and run the programme under gdb.
So, I think this is the way to go. You should be able to figure out even without great knowledge of this kind of thing where the error lies - in your main programme or in a library.
The whole thing is, I guess, so niche that you need to do some work to narrow things down more for others likely to show some interest.
As a guess I would say that a library was updated and your programme does not check on that but struggles (and fails) now.
Peter