I have a few that saved some cost. For example turning of metrics generation for _all_ host headers on NGINX instances serving millions of users. That “feature

I have a few that saved some cost. For example turning of metrics generation for _all_ host headers on NGINX instances serving millions of users. That “feature” cost outages during security attacks and a quite significant SaaS bill, because we were also ingesting the logs of the dying NGINX instances.