:: So, you're back to looking at conditions known to cause
:: crashes: bad UDFs, inflating the TSB beyond the capacity of
:: RAM, external forces (bad hardware, etc.) - or another theory. :-)

Well I don't have UDFs, and hardware is fine. So I suspect the TSB... BUT if
the TSB gets too big, it should still just kill the user - not let the WHOLE
server die. Im sorry but its EXTREMELY poor that a single connection can
easily kill the WHOLE server just by loading up a large transaction. We had
this problem earlier when a transaction ran amuck. This is a "Security" flaw
even. Give me any FB server and watch me crash it in just a minute...