Panic (rwlock error)

Hi,

My cobalt qube2 is panicing quite often (every few days). But it does  
not reboot. It gets stuck in syncing disks.
Here is a capture of the last panic:

Oct 16 18:31:56 qube2 sshd[1612]: fatal: Timeout before authentication  
for xxx.xxx.xx.xxx
Reader / writer lock error: rw_vector_enter: locking against myself

lock address : 0x00000000c005a65c
current cpu  :                  0
current lwp  : 0x0000000084ed4d60
owner/count  : 0x0000000084ed4d60 flags    : 0x0000000000000004

panic: lock error
Begin traceback...
pid -869090792 not found
End traceback...
syncing disks...


At this point, the console is not responsive (but I access it through  
ssh->screen->tip; I tried ~~B)

Does this ring any bell ?
Any thing I should mention about my setup ?

Regards,

Rémi Zara

Some infos:
uname -v = NetBSD 4.99.72 (GENERIC) #0: Thu Sep 18 07:49:27 PDT 2008  
builds [ a t ] wb30:/home/builds/ab/HEAD/cobalt/200809180002Z-obj/home/ 
builds/ab/HEAD/src/sys/arch/cobalt/compile/GENERIC
The box has 256MB of ram.
The disks is 250GB, journaled UFS
Activity:
   postgresql buildfarm (CVS+compiled+tests)
   nfs server for other postgresql buildfarm
   regular large rsync over ssh