otherwise we run with an undef $tid when processing new replies to existing threads.

this wasn't a good idea as the mbox contains sensitive data (IP addresses, hostnames and whatnot). Not good to publish on the web. Reminded by Stefan

mexp is the slowest part of the pipeline. Try to speed it up by paralellizing it. this adds a `pe' scripts that sits in front on N mexp childs and dispatch threads to them, while still outputting the unchanged lines for mkindex. Unscientific testing has shown that this effectively reduces the time for a full export, even if not linearly. (4 jobs cut the time in half, 8 jobs was just barely faster)