If you trust Rule #1 and plan around it you will be ok. It might be the hard drive, the CPU, the memory, the nic card or some other part in the server, but it will fail. My experience is that drives die the quickest and anything older than two years is asking for trouble.
So, implement a great backup strategy that not only has an offsite implementation but a clear and written strategy on how to get things working again.
Not true. I've got battery backups on every one of my machines and on my server rack. I'm glad I do because at least once or twice a quarter my power is out for more than 30 minutes. If not for the batteries, I could never survive. For me, power outages have been the only time that my machines have ever gone down. I bought batteries with interfaces that will shut down my machines safely for me so if they run out of life, everything is ready to go when the power comes back up.
I have my primary servers at home in London in a back bedroom (!) and power problems are on the whole very infrequent and minor. I have seen nearly the same level of power problems in top-quality co-lo. Physical access is valuable to me and security is not an issue (I wanna watch a con try to flog a SPARC1 (20MHz?) and a 486DX down the pub!).
Home hosting suffers from heat (over 30C and you are getting into trouble) and being on the end of a long thin wire from the rest of the Net (for example I am getting 6% packet loss at the moment).
To balance these disadvantages out out I have hosting (eg of DNS secondary and site mirrors) in co-lo, spread out over the world.
(Actually, when I was one of the UK's first ISPs I *did* run it from home, and power was the least of my troubles!)