Data redundancy
Data - redundant, diverse systems. Key data feeds have no common points so it would take two idiots with excavators to take the data out.
I'm sorry to trot out one of my old stories here, but I'll emphasize that assuring no common failure areas (not points) is very, very, hard and expensive.
If you have two data feeds into a location, you can ask your supplier for redundancy and/or resiliency, but what you need is (jargon term here) a specified degree of separacy all the way between data source and data delivery point. Organisations that need or want this (large financial institutions, critical national infrastructure, etc) are versed in requiring that data feeds are physically routed with a minimum distance between the two routes for the whole length of the route. If you are dealing with good old-fashioned physical wires, it can be relatively easily achievable, but once you get into the wonders of multiplexing and virtual services (like MPLS, Metro-Ethernet, and OTN services) you get into a whole new ball game, as you need to assure separacy at all levels of the networking stack. Buying your service from two different suppliers does not, for example, guarantee that they will route their services along separate cable ducts, as it is not unlikely they buy fibre from third parties, and both can use the same third party. I won't go into detail, as it was part of my boring daily job, but it is expensive to achieve. One of the complications is that networks change (additions, upgrades, faults), so a physical path can actually change from day to day, so what starts out as being separate can, by natural attrition (e,g, one path re-routing round a fault) end up sharing a path.
Now for the story.
A customer (who shall remain nameless) specified they wanted a particular degree of separacy for the data entering a location. In this case, it was acceptable to them that two fibre paths were used that routed for part of their length along opposite sides of a dual carriageway. So far so good.
The local council decided to apply a non-default speed limit to the dual carriageway, so, of course had to set up the relevant speed-limit signage. So out went the earth-moving equipment to dig holes for the signs each side of the dual carriageway...I think you can work out what happened.
I don't know if it was a failure of the local council's GIS*, or 'idiots with excavators' not following instructions, but the end result was the same. In my view, you should always be prepared for an outage, even if you have taken great (and expensive) lengths to not expect one. Reality has a tendency to find unexpected flaws.
*I have a good few examples of what the GIS said and what was actually installed in the ground being very, very different.