I remember a story about someone who did something similar with a server that kept hanging. They rigged up a second computer to ping it over the local network and if there was no response for a certain amount of time, the computer would eject its CD-ROM tray which had been lined up neatly with the reset button on the server.
Since it couldn't eject fully, it then retracted, having rebooted the server.
I assume that was a temporary fix... and it was probably a Windows server tbh.
The closest I've done is having a job run every 12 hours checking if a process was over a certain memory usage (memory leak) and restarting it if it was. That was also Windows, but the same thing on Linux wouldn't have been difficult... not that the Linux servers ever had that problem.