Hello @romain! Welcome to the community.
Yes, you understand that correctly. For now, I would simply copy example input data and then paste it in a Function-Node after the Start-Node. Best to add a NoOp-Node after the Function one to be able to reference data from that node (to have one fixed start-point). You can then always when you want to test, simply activate/deactivate the Function-Node.
Thanks a lot for the input. It is very helpful! It is planned to add something similar to what you described to make the process easier in the future. As like you realized it is currently not optimal. Sadly have however no estimate yet as the workaround (even though it is not great) is not too complicated and I want to make sure that the functionality we implement here can also be used for another realted feature that is planned.