Parallelvm and bigfork, both, when run individually always pass on my system. However when they’re run as part of vm-stability, either (or sometimes both) of them fail. Specifically:

parallelvm prints SUCCESS on screen, but it says it failed in the output summary.
bigfork fails saying [n] failures.

This inconsistency exists right from 3.2. How do I go about debugging these failures? Where should I start?

I would suggest editing the configuration you’re running with. Try up to 32 CPUs and as little as 1M of RAM. Then you should hopefully be able to replicate the error running these programs through the shell and poke around with GDB. However, I don’t believe GDB will help you much - this is very likely a synchronization issue to do with your swapping code. Try adding more assertions in functions like vm_fault, as_copy, etc. that ensure what you believe about the pagetable entry and coremap entry you’re working with are really true. E.g. Is this page in memory before I start copying? What about after? Does the coremap entry still belong to me?