Complex issue I guess as we first need long-term and high-quality sub-daily reference dataset such as station observations. So not feasible for many areas or many variables. But this has been probably done at least locally for some stations over Europe for precipitation. I have no reference in mind.
But model output is 6hourly and that needed for flash flood modelling....?
I agree that sub-hourly frequency is needed for flash flood modelling. In that case, especially in not-well observed areas, this is an open issue.
