I believe AF-628 is a SEGV on the program counter, which would explain why our stack trace stops at the signal boundary. Dumping the registers from the ucontext should confirm this, but will likely get us no closer to the root cause of the bug.

This may also be generally useful for future debugging.