When your DS3 goes down and all your customers start calling in about not being able to access the Internet, and your upstream provider doesn't give you any ETA on a repair, the fight-or-flight syndrome kicks in and you find yourself Routing Under Pressure.
Quickly, you have to find a way to redirect all the traffic through your backhauled T1 to another POP so that the users can at least check their email. Now, you start the hard part -- breaking that ancient, unused, but routable Class C (which you haven't touched in six years) into appropriate subnets and turning on static NAT so that you can get the traffic out of the stranded network. You do the math quickly -- was that a /26 or a /28 subnet that you're using? Oh, no! I hope we didn't just change the default route from the 4700, since we're seventy miles away and if we lose connectivity to the command line interface, someone will have to drive out there and reset the bugger. Phew, everything seems okay and we're able to pass traffic to at least the remote access servers. Back to the subnet; was .32 the first usable address or was it the network itself? What was the broadcast again? You're wrong -- it has to be odd, so it can't be 48, wait a second, did we just assign an alias or a primary address on the 2501? Does anyone know what version of IOS is on the 2611? Can it handle nat? It can't? Well, drop those routes we just added and point everything at... you made a loop between the Linux webcache and the 2501? In the good way or the bad way? Okay, so what does that buy us? Great! Now the users are able to surf, even if they're all limited to 1/45 of the usual bandwidth. If we're lucky this won't last all day. Anyone got the trouble ticket number? Back to that subnet calculation. Wait a minute, the traffic isn't coming back on the right pipe because Sprint is pointing those addresses at the downed DS3. Should we try to turn on NAT now, so that all the other protocols will work? I agree, let's not break anything.
Anyone up for a game of Quake?
(Note that none of this situation doesn't matter if you're running BGP on your exterior routers, but until today you've never really thought it would be necessary. Hindsight is 20/20, eh?)
(Note 2: My boss thinks "Routing In Pressure" would be a more appropriate name, mostly because of the acronym.)