KDE 4 processes get into unkillable state

Hi,

sorry for the crosspost, but it's relevant for both lists.
After a random amount of time, or more to the point - after a certain network 
condition, KDE4 processes can get into an unkillable state.

procstat -t:
  PID    TID COMM             TDNAME           CPU  PRI STATE   WCHAN    
 2237 100077 klauncher        initial thread     1  128 lock    *unp_mtx  

procstat -k:
  PID    TID COMM             TDNAME           KSTACK                       
 2237 100077 klauncher        initial thread   mi_switch turnstile_wait 
_mtx_lock_sleep uipc_peeraddr kern_getpeername getpeername syscall 
Xint0x80_syscall 

WARNING: running procstat -f on the process hangs the entire machine, no panic 
generated.

It's hard for me to generate a test case for it, but I've seen it in various 
KDE4 processes, at least in konqueror, klauncher, kio_http and kontact. Also 
pulseaudio when I was using that sound backend briefly.

Anyone have an idea whether this is KDE4 specific or (what my hunch is at the 
moment) what scenario can create this hang in getpeername?

If anyone is interested, I will try to collect more info the next time I see a 
process and if so, please let me know what info is useful.
-- 
Mel