Good idea Victor. After a couple hours the backends got 150MB large each and the whole thing put the server down. So what I'll do is make sure the backends die after serving a couple thousand of requests. This should make me safe from such errors. I can't just ignore a 100% increase in performance.
Regards
