The trouble is that there are so many security issues in all software products these days, that it’s hard to distinguish them.
This week’s NGINX and Ruby vulnerabilities seem bad, and have received a lot of press attention, but AFAIK there are no published proofs-of-concept outside of lab environments (yet). There are many other recent vulnerabilities in Discourse and other packages which have presented far more risk for Discourse communities.
When we update something in the base image which is known to be critical for Discourse (e.g. the libheic updates a few weeks ago), we aim to cut a core version around the same time, so it’s all covered through the same notification system. Most people don’t come to Meta, so releases.discourse.org, the admin dashboard, and the automatic notification emails are the main way we distribute that information.
My advice would be that everyone should be applying every update as soon as possible after they receive a notification. Every one should be considered critical. The speed of vulnerability discovery and potential exploitation is so fast these days - waiting is not an option.