What can be done to increase the reliability and timeliness of the daily
supscan? It seems to fail about one time in five, which means we wait for
either the next day's scan, or for someone to notice and manually re-run it.

It would also help immensely if the scan's termination time could be
reliably be determined - I start my daily builds "sometime afterward" and
when the scan "takes too long" my build systems end up building the same
sources as the day before.

Naturally, the slower the hardware, the more important it is to waste as
little time as possible...

	Erik <fair@clock.org>