> Perhaps we need a new syscall to *really* kill a process, even if it's
> in a state where normal signal delivery does not happen.

Urk -- when processes are unkillable there's a reason.  The reason is
sometimes hardware and sometimes a kernel bug.  Neither will be fixed
by removing the process, and doing so will mislead administrators who
don't understand what is going on.

A system with a long term unkillable process is headed for a reboot,
possibly via a crash dump to collect the evidence, but a reboot
nevertheless. :-)

Giles