Thanks for the PR @harry671003 . The code looks good to me. Just a small suggestion: the last_over_time function doesn't explicitly handle the case where there are no samples at all. While this doesn't cause test failures due to earlier checks, adding an explicit condition for empty input could improve readability and make the logic more self-contained. WDYT?

Thanks for the PR @harry671003 . The code looks good to me. Just a small suggestion: the last_over_time function doesn't explicitly handle the case where there are no samples at all. While this doesn't cause test failures due to earlier checks, adding an explicit condition for empty input could improve readability and make the logic more self-contained. WDYT?

If there aren't any samples at all, the code in question isn't even reached because the range selector doesn't select anything. Right?

In different news, the case that triggered the problem does have samples in the range, namely histograms with a negative timestamp.

Maybe I'm missing something and there is room for improvement, but let's first merge the solution as is to stop the bleeding.