See resolution of Support Request with Nicholas

https://support.atlassian.com/browse/CSP-10538

While the RSS feeds do reach the DAO caching layer and the database is minimally impacted, the CPU utilization of the VM is huge - directly proportional to the "items" parameter of the feed request.

I think the workaround is to install one's own caching at a front controller HTTP like Squid and/or to filter GET requests and limit the size of the items value.

I think that product team needs to solve this caching matter in the product - at least for anonymous GET requests it should be very easy to add this to the existing caching infrastructure. I think that the Admin Console and Confluence configuration needs to have an optional maximum items parameter so that the CPU load can be somewhat insulated.