I recently came across some scanned notes from a TechReady conference from (I think) 2007. Anyway, the topic is fascinating. Recovery-oriented computing is a method for developing reliable Internet services. The idea is to reduce the effects of harmful events, like Bohr- and Heisenbugs, amongst other scenarios:
The idea is to automate the recovery process depicted above. This leads to some interesting side-scenarios, like obtaining database connection strings, automating database creation, and application configuration.
Some of these things are relatively easy to automate these days. What I found useful was to apply a similar philosophy to Interact client applications, which run on the desktop and don't benefit from technologies like Red Dog and SCCM.