Right now it's more of "can we" not "we must."
The general use case is that we have some list of tables we need to process in a Flow. Task 1 queries for the list of tables, Task 2 gets mapped to this list. Then we run Task 1 again and if it returns a non-empty list, we pass that to another run of the same Flow. We're just looking for other options to do this processing and maybe keep it all in one Flow.