Sigh. As may be obvious to anybody unfortunate enough to look at the link, I seem to have gotten my core files confused. I have a bunch of db_stat -CA done properly, but the backtrace for this may be lost to the bit bucket. I'll poke around my systems and see what I can find, but I think it might be nothing. A syslog of "waiting for 1027 threads to terminate" might be the only other durable evidence.