The ongoing DDoS had shifted gears and they're attacking database intense functions of the stats from thousands of IPs making it hard to filter.
Its causing *stats side* database replication lag, which is playing havoc with random hashrate calcs.
Pool side (ie: the important stuff like earnings) are not affected.
The attackers goal seems to be to stir up confusion. Judging by these posts its unfortunately working. :-/
What I did on one of my sites running nginx and php was to use xcache to store the results of database calls for a period of time. I had wrapper functions for any database info I needed, and would cache the result for 30 seconds up to 15 minutes depending on what it was. This way even if someone did an endless spam of page reloads, it'd only affect the web server and the database server wouldn't even notice. Maybe something like that would help.
There is already pretty robust caching in place... however it doesn't help when the attack is specifically designed against the Eligius stats and each request is looking for new information. These attackers are pretty determined.