PowerMac G4 unstable during pbulk

Hi,
When running a bulkbuild at some point (some hours or days) cpu usage drop
to 0%. System is still responsive but it is in a weird state.I've tried to
kill all process linked to pbulk and restart it but I'm always unable to
kill one fetch process even as root with "kill -9". I can't even reboot
because it seems nothing can kill this process.


nbqc1-mppc1# ps auxww | grep -v grep | grep fetch
root     6013  0.0  0.3 10916  2952 pts/1- D    Mon02AM 0:00.10
/usr/pkg/bin/fetch -o slib-3b5.tar.gz
http://swissnet.ai.mit.edu/ftpdir/scm/slib-3b5.tar.gz
nbqc1-mppc1# kill -9 6013
nbqc1-mppc1# ps auxww | grep -v grep | grep fetch
root     6013  0.0  0.3 10916  2952 pts/1- D    Mon02AM 0:00.10
/usr/pkg/bin/fetch -o slib-3b5.tar.gz
http://swissnet.ai.mit.edu/ftpdir/scm/slib-3b5.tar.gz


Restarting pbulk while have this "unkillable" fetch process ends in pbulk
hangs ( cpu 0%) on the first package.

Furthermore, it seems some (or all) off my cronjob hangs :

nbqc1-mppc1# ps auxww| grep cron | wc -l
      24

At least those one are killable.
I have nothing of interest in /var/log/messages...
Problem was there with NetBSD 8.0, 8.1 and 9.0. I'm also running netbsd on
my ibook G4 as well as emulated qemu host and none of those have this issue.

On one hand I tend to think it might be an hardware issue with my PowerMac
( failing hard drive or ram)
However on the other hand i'm asking if there might be a generic issue
related to this hardware.( Something kernel/module/driver related that
might explain the unkillable process)
Has anybody experienced such issue with a PoweMac G4 ( or any other
hardware) ?