Like Morgan Tiley in the first comment, this was also reminding me of Google's approach to distributed systems. There, the collections of systems are big enough that you will have a chaos monkey just from hardware failures, so you have to build the system to deal with that -- at which point, they famously asked why use expensive high-reliability hardware when the cheap stuff is vastly cheaper and only less-than-vastly less reliable? Seems to work well for them. And, on the face of it, hardware failures are much less friendly than an artificial Chaos Monkey that you can simply reboot from. It's definitely an interesting approach to include a mild one voluntarily, though!

@Steve, @Alanstorm: The actual link is http://www.randsinrepose.com/archives/2004/01/26/beta_is_dead.html -- Alan's link has the trailing ')' attached.

The product that my team is developing doesn't have a UI in the sense that most people think of; it's a C++ library, so its user interface gets called an API instead. Nonetheless, issues of usability apply quite strongly here, as well. Are books like Rocket Surgery Made Easy useful for this sort of thing? Any good recommendations for books and references on how to make APIs more usable?