Recurring Patterns Across Major Outages
My read after enough of these reports is that failure has a small vocabulary. You expect infinite variety and you get, instead, maybe a dozen recurring shapes wearing different costumes. A saturating feedback loop. A dependency everyone forgot was shared. A recovery step that only worked in theory. The trigger is the costume and it changes every time; the shape underneath is old. This is good news, oddly. A field with infinite failure modes would be hopeless. A field with a dozen is one you can actually study, memorize, and hunt for. Learn the shapes and you stop being surprised by novelty that isn't novel.
Knowledge check
What is common-mode failure, and why does adding more redundant copies of a component sometimes fail to protect against it?
Show the answer
Common-mode failure is when multiple components that were supposed to fail independently fail together because they share a hidden common cause, such as the same configuration, the same dependency, the same software version, or the same power source. Redundancy only helps against independent failures; if all your copies share the flaw or the dependency, one common cause takes them all down at once, and the redundancy was an illusion.