Quoting Chad Walstrom <chewie at wookimus.net>:

> Josh Welch <josh at joshwelch.com>  wrote:
>> This is why I actually like rebooting servers every so often,
>> preferably in a controlled fashion when you have some available down
>> time.
>
> Actually, we test reboot servers before patches that require reboots
> (i.e. new kernel).  That way, if the server fails to recover from the
> test reboot, there are bigger problems to solve (i.e. disk failures).
> You don't end up wasting hours applying patches only to have the
> server never come back to life.
>

That's a hell of a sensible policy.

Josh