We're a pretty small shop, with 4 production servers. We've been running one server for 7 years, and the other 3 for the last 2.5 years.

So far <crossing fingers> we haven't had any serious hardware failures. We had one motherboard lose a ram slot, and another box that stopped powering on after a power failure or APC reboot. We were able to fix both of those during scheduled maintenance.

If something did break, there has to be a replacement on-site, or the downtime would get quite long while we located a spare. We found the most economical way of ensuring a spare was to ask our datacenter what servers they use and order the same (Supermicro).