We’ve made some changes to our architecture that have significantly speeded up the execution of flows. 
You should especially see better performance where you have several filtered tasks between other messages.
The platform is still designed to be highly distributed and fault tolerant.