I think the paradigm is a lot more flexible but for me it would boil down to that flows have properties that tasks don;t (schedules, storage, FlowRunner). I would personally say keep stuff as tasks until you need one of those properties.
Also, flows that group tasks will support retrying together if not already. A common request in Prefect 1 was to retry upstream tasks if a downstream one fails.
I think it also depends what you need visibility for in the UI and you can think about how you would filter for failed flows/tasks