The systems that ran all night for nobody
Sandbox and development systems ran around the clock, weekends included. We put them on a schedule that stops them when the team goes home and starts them before anyone logs in, and right-sized what stayed on. The bill dropped and nobody noticed a difference.
The certificate that renews itself
A wildcard certificate on the load balancers had been renewed by hand every year, and one year it nearly was not. We moved it to the cloud certificate service with DNS validation. It renews on its own now, and the change board got a one-page write-up instead of an annual scramble.
The disk alarm at six in the morning
A storage alarm fired on the HANA shared volume before the business day started. The culprit was a folder of database log backups from years earlier that nobody had touched. We archived it, cleared it and had the space back before the first user logged in.