I'm not sure if it's about the implementation rather than the vast reduction of complexity/boilerplate to attain fault-tolerant distributed systems compared to other solutions.