The code around management of sort/distinct/limit should be refactored to allow for easily adding new stages as well as for planning and optimizing the process from a high level view. We could introduce a pipeline planner that has the ability to make optimizations such as: if we know we are doing a sort followed by a limit, we can restrict the sorting stage to only store and sort the N limited items instead of sorting all items in the results.

Making time for this project would be a great first step before doing Group By. It would also make it easier for new features such as #6759