I remember burning in servers. I've burned in drives before. It was a 24 hour process designed to find the flaws before you put the device in service. Did you ever run across anyone who thought the server "ran better" after burn-in? I dunno. The whole thing is weird.
In the early days of hard drives, I worked for Seagate. Part of my job then was involved in the burn-in process. In those days, we'd make the drives seek continuously for 48 hours, and a few percent would always fail and be scrapped. We improved on the process though, and I developed the code we ran on the drives themselves during burn-in. We converted it from a simple pass/fail burn-in, to a flaw-mapping / media characterisation phase - so that, for the drives that survived burn-in, we would have a wealth of info stored on the drive itself. This allowed tuning of the read-back parameters later, which saved the company a shit-ton of later testing and flaw-mapping. The genius part was that all this manufacturing phase needed was a power supply for those units. Saved millions in test equipment costs.