I guess now is the time to open my mouth and try to place my foot inside.
 
The various  crashes causes by the  installation of 8803  are due to a  bug in
LSVBITFD. I must have been completely drunk the day I wrote this.
 
I was using  a wraparound stack buffer  which was wraparound only  in my mind,
since I had forgotten a few  instructions in the relevant places. Fortunately,
the stack  was oversized  (1 entry per  node, which is  the maximum  number of
entries it  can receive  in theory), and  we never had  any problem.  With the
addition of the HEARN-FRMOP22 line, the  network complexity grew and the stack
was overflowed.
 
Anyway I've replaced all this with a direct-access array of flags coupled with
a chained  list which solved the  problem and had the  side-effect of speeding
the  program  by 20-25%.  I  still  can't understand  how  I  could have  used
something as slow  as this "wraparound" heap which  didn't prevent duplicates,
but nevermind.
 
The  updated  version  is  being  tested at  CEARN,  BNANDP11  and  HEARN.  If
everything goes fine,  I will send a FIX15M4 shipment  so that Hans-Ulrich may
send the 8803 files to the States. Since FRULM11 is down, you're likely to get
this mail a long time after the shipment. Sigh.
 
  Eric