large number of SIGHUP cycles, these would have run the postmaster out
of memory.  Noted while testing memory-leak scenario in postgresql.conf
configuration-change-printing patch.