According to Google's own HDD failure report, during their first years of life, hard disks tend to fail less often when utilized less. Therefore, if you can take some load off that single hard disk, it could potentially extend its lifespan.
One way to do this would be moving data that is often used to a ramdisk, ie utilizing a tmpfs on a *nix system and rsyncing data back and forth (an implementation of such a practice is demonstrated in the Anything Sync Daemon).
A temperature of 40-45°C also seems to help.
A question that might also interest you is:
However, all hard disks fail at some point in time and nothing will prevent this. The only certain way to add reliability is by adding more storage devices to the equation.
As of having a bad power supply, you should replace it at once. A bad PSU can potentially destroy one or many components, including a hard disk board, rendering it unusable.