Now that vicious eject is disabled, we've valuable evidence showing what
causes slot hangs.  I'm working on diagnosing this, but in the meanwhile
here is what the core file says in gdb about the process state:

(gdb) bt
#0  0x401751fe in read () from /lib/tls/libc.so.6
#1  0x08055099 in doRead (asock=5) at socket.c:1002
#2  0x08055925 in readFromClient () at socket.c:938
#3  0x0804dc48 in input () at input.c:152
#4  0x080503b5 in main (argc=Variable "argc" is not available.
) at main.c:437

And lsof shows the TCP connection down but the UDP connection still up:

COMMAND   PID  USER   FD   TYPE  DEVICE    SIZE       NODE NAME
ntserv  20876 quozl    0u  sock     0,4            4364912 can't identify protocol
ntserv  20876 quozl    1w   REG     8,1   35566     130991 /usr/games/continuum/var/ERRORS
ntserv  20876 quozl    2w   REG     8,1   35566     130991 /usr/games/continuum/var/ERRORS
ntserv  20876 quozl    3w   REG     8,1   35566     130991 /usr/games/continuum/var/ERRORS
ntserv  20876 quozl    4w   REG     8,1   16639     130994 /usr/games/continuum/var/god.LOG
ntserv  20876 quozl    5u  IPv4 4364929                UDP 208.20.202.42:39737->70.172.223.17:12415 

And strace shows continuous calls to read the UDP socket being
interrupted by SIGALRM (the daemon synchronisation signal):

read(5, 0xbfffb8c0, 16384)              = ? ERESTARTSYS (To be restarted)
--- SIGALRM (Alarm clock) @ 0 (0) ---
sigreturn()                             = ? (mask now [CHLD])
read(5, 0xbfffb8c0, 16384)              = ? ERESTARTSYS (To be restarted)
--- SIGALRM (Alarm clock) @ 0 (0) ---
sigreturn()                             = ? (mask now [CHLD])

-- 
James Cameron    mailto:quozl at us.netrek.org     http://quozl.netrek.org/