Thanks for the report. Bert and I have looked into it and I'm hopeful
that it is now sorted out. It looks like it might have been an issue
with PHP/mod-fastcgi, presumably a resource leak of some kind that
resulted in a problem after 241 days of uptime.
Ken