I've setup a central rsyslog server that is writing events to a MySQL database.
Looking through the events one sees all kinds of events that are probably unnecessarily captured.
On RHEL/CentOS systems I see and am thinking of discarding the following (in rsyslog.conf format):
:msg, contains, "(root) CMD (run-parts /etc/cron.hourly)" ~
:msg, contains, "running program /usr/sbin/rhn_check" ~
What other types of messages are candidates for discarding?
This entirely depends on your own preferences. After all, there's nothing in the logs on the remote syslog server that are vital to keep the original servers running.
I would personally not discard much, since you never know what you might need to troubleshoot or to perform forensics after a break-in. Storage is relatively cheap, and logs compress very well, so I would suggest keeping as much as you can, and determine a period after which you purge old logs. I'm not sure how you would compress the logs when they're stored in MySQL.
If you do want to filter logs, candidates for removal would be the messages with the least value for later analysis, such as the hourly recurring ones you already mentioned.