The Limits of Safety : Organizations, Accidents, and Nuclear Weapons
It's the height of the Cuban Missile Crisis. Bombers sit fully armed at the end of the runway, awaiting the order to penetrate Russian airspace. At night, a sentry sights a man climbing over the fence. He shoots at the saboteur and sounds the alarm. Linked alarms go off at several nearby airbases, except one alarm isn't the sabotage alert klaxon, it's the alarm to launch the bombers. Crews pile into their aircraft and the bombers trundle down the taxiways. All of a sudden, lights appear on the runway. It's the base commander in a jeep flashing its headlights. He's called over to the other base and found out the situation and stops the bombers.
The saboteur was a grizzly bear.
This true story (I'm paraphrasing actually, you can read a sample on Google Books) starts a very interesting book analyzing US handling of nuclear weapons during the Cold War. How the Air Force (and Navy) handled The Bomb is used to teach all sorts of aspects of Safety/Reliability Engineering, from High Reliability Theory, to Complex Systems.
For a book primarily aimed at academic audiences, it's pretty readable. True stories of how we almost set the world on fire keep it interesting. There's also plenty of examples of robustness in the system, too. Taleb in the Black Swan mentioned the one group he's come in contact with that had the most robust view of risk was military officers (far better than bank officers).
I'm now working for a new client and the book should prove useful. This client is so anal-retentive about safety they have pee coloration charts above the urinals to tell whether or not you're dehydrated. This isn't for offshore workers, either. This is for downtown officeworkers.
Note: Image and title link lead to Good Reads.