I do know this data is flawed due the absence of now deleted posts but more disturbingly there are no rows in the reviewtasks table for reviewtasktypeid = 2 before July 10th, 2012. I tried to re-invent that data by bringing the rows for votetypeid = 6 in the Votes table in my dataset. That still shows a giant jump, because close votes expired after 4 days. The posthistory table for posthistorytypeid = 10 isn't of much help either.
Am I missing something in my reasoning and therefor in my query? Where did those 47,641 review tasks at the start of recording in the reviewtasks table came from?