In February 2016, Netflix announced it had finished moving its streaming service to AWS. It had taken about seven years. The trigger was a single bad week in August 2008, when a major database corruption left Netflix unable to ship DVDs for three days.
Netflix is the reference story for "moving to the cloud" for two reasons. It didn't lift and shift: it rebuilt the service to suit the cloud. And it treated failure as normal, building tools that broke its own production systems on purpose. Both ideas are now standard practice, which makes it easy to forget how unusual they were.
What they had
Before the move, Netflix ran in its own data centres, with a monolithic application and relational databases that were, in the post's framing, single points of failure. The 2008 corruption showed what that meant: one database problem stopped the business for three days.