Looking back, it was 3, and I had forgotten that the fourth paper I published was somewhat different. The initial referee report was pretty ridiculous, but we got that one published without having to request a different referee.

Some of the most ridiculous referee reports I've gotten back first complained about the paper, then included a citation request to some obscure paper which I can only guess was by the referee. Those seem particularly weird; if you're going to try to use the refereeing process as a way to get yourself more citations, wouldn't you want the paper to get accepted for publication?