Thanks for including the diagrams. They always help!
I suspect there is an issue with fault opening based on the nonplanar fault geometry. On some systems, stdout and stderr get truncated and a simulation ends before it produces the error message. This is more likely when running in parallel. Sometimes just capturing stdout and stderr to a log will enable you to see the error message.
See Nonzero traction error in spontaneous rupture model - #10 by baagaard for a discussion of setting zero tolerances for nonplanar fault geometry.