For periods that are Date ranges (period=range) and that are:
then, if it has been pre-processed before, it should use the existing Week / month or year report.
In my tests, the second request was not using the blob from the first request, as it triggered archiving (and took 20s to return).
So, we should fix the algorithm which will result in better performance and user experiece.
The second request was slow because it was selecting the weekly data and then copying it into the new archive. We could avoid such copy when the range is exactly an existing range, but this won't make much difference.
So, the assumption about range not using existing period was wrong. It already works...