As described in the title.
This is useful to avoid hitting API caps.
For some of the larger datasets we've implemented. Our process has been. 1) iteratively build out the historical dataset
2) switch the query to "last x days" and "combine new results with old"
3) set a scheduled refresh for every xth day
* The feature should be valid even when the user selects date as a column. We are using this feature with date almost all the time and there are some reports we put the date as a column.
* The feature doesn't have to be valid without a time dimension. I don't think any use case without a time dimension.