All sorts of ways. If you can't disrupt services at all, you should have several machines. If one goes down, the others pick up the slack. Serious setups may have machines in several different geographic areas in case one gets hit with an atom bomb. This isn't in your reach and won't help you.
For us normal people with a checkbook, generally what you do is not run windows. That takes care of 99.999% of your problems right off the bat. Next, you may have a watchdog card. This is a piece of hardware that constantly talks to software. If the software stops responding for X seconds, the hardware card knows to reboot the server. There are software-only solutions which work fairly well, but they can fail. Next up, your machine is probally at a colocation facility which has humans running around 24x7x365 doing stuff. When you sign up with the colo facility, you get a pager number or a webpage and you're told how to send a message and how to authenticate yourself. When they get a page, they walk over to your machine and punch reset.
Pkcp