<stuff snipped>
The problem we've been encountering is that every so often (several times a day) the server goes nuts and I end up restarting it. I can watch "top" and see a single python process take all of one CPU time for an extended period.
are you using any RDB? we used to have this problem where suddenly, only one zope thread is active. we were using postgres-6.x. an upgrade to postgres-7.0.x and PoPy, we managed to stop having such problem. dunno whether that has anything to do with the "single active zope thread". we are now considering upgrading to 7.1.x and psycopg. on a devel box, postgres7.1.2 and psycopg really flies
If it ends up in the ZEO client (as I suspect it will), I can use the front-end proxy to route requests to other ZEO clients while I restart the one that's spinning. I can also restart the spinning ZEO client quickly.
yes, ZEO did save our backend when we were having this intermittent trouble.