As can be seen in the screenshot, the total number of unread notifications on Wikidata is given to me as 100, but the number for Q229883 (the item with the highest activity) is given as 386, which indicates that the 100 is probably wrong.
My suspicion is that the 386 may also be off - even by one or two orders of magnitude. The reason I am suspecting this is that a few days ago, that number stood at about 2000, and I have since not gone and looked at more than a few single relevant diffs, while notifications relating to Q229883 have kept coming in in great numbers due to a series of WIDAR edits by another user.
I took some more screenshots over the last few days in case past numbers become relevant.



