As systems become ever more complex, more and more “layers of truth” are added to the system (search indices, caches, read replicas, etc). How do we manage them when things go awry?
This is a talk about preventing cascading failures by treating failure as a normal mode of operation.
