This feature exists but is very underdeveloped. We basically just duplicate the RUME with an altered TimeFrame and run on that. I think this is probably problematic given the complicated relationship between ADRIOs and the RUME TimeFrame, and is unlikely to work in all but very simple cases.
A better approach might be to do all of the data fetching for the larger RUME (which improves memory-cache availability) and then subset all time-series data if needed at run time. Multiple runs should then should be able to re-use the fetched data efficiently, even if the time frame changes.