I mentioned it before, but I'll just say it again. The biggest problem when you provide web based solutions is when something screws up on the network backbone where your server is sitting on.
The process of proving that it's not your application, your server or your network equipment to finally pin pointing that your network provider is the cause of the problem is just nerve wrecking!
Worse still if you have your client breathing down your neck asking what's wrong, why the system is down even though we have multiple redundancies?
Well... we do have a fail over cluster... but not a remote server duplicate sitting on another network of course.