Comment by refurb

This seems like a super important observation.

If the number of dataset saw a massive jump (50%), then back down a week later, that seems more like the correction of an error.