There are actually multiple optimizations to Flux already underway. The language runtime can "push down" some of these operations to the storage layer where they can be performed more efficiently.