Skype's outage last week created a whole range of rumours. Was it a DDoS? An attack against a known vulnerability? Some bug?

It seems that the outage was caused as an unanticipated consequence of Microsoft's path Tuesday. As a major patch was rolled out, millions of computers rebooted. As skype is a peer-to-peer network it did not suffer from a single point of failure. But a balancing algorithm that allocates resources on the network suffered a "death by a thousand papercuts". The rebooting machines caused an instability that continued to grow until it exceeded skype network's ability to adjust.

Patches always carry risk. They introduce change into the environment and can conflict with existing software. Testing patches prior to deployment can minimize the risk somewhat. But one thing that this skype outage shows is that in today's massively distributed networks there are dependencies and potential instabilities that cannot be seen in advance. Massively distributed systems reduce the risk of a single-point of failure but they also are susceptible to system-wide effects that may not be predictable. If planning won't help, then being prepared for a rapid response is the next best thing.