Since the source graph gets wonky immediately before the crop, there’s not really enough here to compare against from the before-times (last week).

But what is here looks like a very large problem on the comment front, with each peak being lower than the last and and the latter two nadirs following the same pattern. The presumed “small number of power users” were having a noticeable impact more than 48 hours earlier. (and, hey … good on them for using 0 as the axis)