Comment 5 for bug 1248181

But you're not getting a wrong context, you're getting nested signals (presumably in the real case SIGPROF is the first signal, because the sigaction looks like it'd mask prof, but prof not mask anything else.)

I'm not sure you need to match the program counter in the context with anything?
Is SBCL assuming the os_context_t refers to user code, and is just absolutely unprepared to handle nested signals? If that's the case, it needs to ensure they don't happen (not just on SunOS, but on any unix)