Another reason why improving the worst-case in individual components of a distributed system improves the best-case overall is that most requests end up seeing a worst-case latency somewhere in the system. Gil Tene has a good blog post explaining why.
Another reason why improving the worst-case in individual components of a distributed system improves the best-case overall is that most requests end up seeing a worst-case latency somewhere in the system. Gil Tene has a good blog post explaining why.