I've got the report, it's AA587, and available from the NTSB site.(It was Nov 12/2001, report AR04/04) Many things contributed (of course) and to put it simply (and in my own words) included aggressive and repeated rudder inputs, the tendency of this FO to do just that in similar situations, the AA training which encouraged the use of plenty of rudder for wake turbulence upsets, the simulator used for this training did not have the same control behaviour or response rate as the A300-600, and the rudder control on this aircraft had, to term it simply, a type of "power steering" that didn't change the feedback through the speed range, nor reduce the rudder deflection applied with any given foot force, with increasing speed.
Doublets and reversals, definitely.
Still, it does seem strange to me that a tail can be torn off this way inside the envelope. I guess it did to the NTSB too, coz I recall they spent rather a long time testing composite materials for strength/fatigue etc, I think at an automotive plant, before the report was issued.
There is also a video, and a computer animation.