updated to use FLOGFILE= and to use .bz2-compressed event pickles.
The intention for production nodes is to use the "log gatherer" feature to automatically send all log events to the central gatherer.
Eventually, my plan is to have one logfile to which major events are written, and then have minor events only get written (to the pickle) when things go wrong. But we're only halfway there, and I fear that we've got the wrong half done so far, resulting in fewer log messages recorded rather than too many.
