was that applying maximum and opposite control response even below maneuvering speed can cause structural failure. Pilots are taught that when you're flying at or below maneuvering speed, you can apply full control deflection without doing damage. Normally, the plane will stall before enough Gs are pulled to do damage.

However, that has proven not to be the case. It is generally true if you're talking about positive Gs in the pitch axis. However, it's possible to exceed the negative G limit in pitch at maneuvering speed. It turns out that you can exceed the limits on other components like the vertical stabilizer if you apply full control inputs. The Flight Data Recorder (FDR) shows the copilot (who was flying the plane) was applying full left and right rudder inputs in response to the turbulence from a preceeding plane. The engineers knew this but the pilots didn't.

A few months after this accident, the NTSB issued an advisory to pilots of many types of airliners warning them against applying heavy opposite rudder inputs.

This safety recommendation letter addresses an industry-wide safety issue involving omissions in pilot training on transport-category airplanes. Specifically, the National Transportation Safety Board has learned that many pilot training programs do not include information about the structural certification requirements for the rudder and vertical stabilizer on transport-category airplanes. Further, the Safety Board has learned that sequential full opposite rudder inputs (sometimes colloquially referred to as “rudder reversals”)—even at speeds below the design maneuvering speed1—may result in structural loads that exceed those addressed by the requirements. In fact, pilots may have the impression that the rudder limiter systems installed on most transport-category airplanes, which limit rudder travel as airspeed increases to prevent a single full rudder input from overloading the structure, also prevent sequential full opposite rudder deflections from damaging the structure. However, the structural certification requirements for transport-category airplanes do not take such maneuvers into account; therefore, such sequential opposite rudder inputs, even when a rudder limiter is in effect, can produce loads higher than those required for certification and that may exceed the structural capabilities of the aircraft.

The advisory discusses the events of the Flight 587 crash. Here are a couple of pertinent paragraphs:

Before the separation of the vertical stabilizer and rudder, flight 587 twice experienced turbulence consistent with encountering wake vortices from a Boeing 747 that departed JFK ahead of the accident aircraft. The two airplanes were separated by about 5 (statute) miles and 90 seconds at the time of the vortex encounters. During and shortly after the second encounter, the flight data recorder (FDR) on the accident aircraft recorded several large rudder movements (and corresponding pedal movements) to full or nearly full available rudder deflection in one direction followed by full or nearly full available rudder deflection in the opposite direction. The subsequent loss of reliable rudder position data is consistent with the vertical stabilizer separating from the airplane. The cause of the rudder movements is still under investigation. Among the potential causes being examined are rudder system malfunction, as well as flight crew action.

Preliminary calculations by Safety Board and Airbus engineers show that large sideloads were likely present on the vertical stabilizer and rudder at the time they separated from the airplane. Calculations and simulations show that, at the time of the separation, the airplane was in an 8° to 10° airplane nose-left sideslip while the rudder was deflected 9.5° to the right. Airbus engineers have determined that this combination of local nose-left sideslip on the vertical stabilizer and right rudder deflection produced air loads on the vertical stabilizer that could exceed the airplane’s design loads. The Board notes that, at the time the vertical stabilizer and rudder separated from the airplane, the airplane was flying at 255 knots indicated airspeed (KIAS), which is significantly below the airplane’s design maneuvering speed of 273 KIAS.

Immediately following 9/11, quite a few pilots said that if someone attempted to take over their plane, they were going to bounce them around the cabin like a rock in a can. They were going to use heavy control inputs to do this. It turns out that the flight simulators the pilots train on don't adequately model excessive G loads due to full control deflection, so they just didn't know how dangerous that could be.

As to the repair, I really don't know for sure. One thing I've learned over the years is that an attempt to reenforce one part of a structure can often increase the stresses at another point. Looking at the pictures, it looks like the patched attachment broke off right above the repair while the other one appears to have peeled off. The other two attachments seemed to break off at the lugs. There really isn't enough to tell from the photos as to which attachment broke first, which probably caused a chain reaction. Believe me, NASA, the NTSB, and many other aerospace players are investigating this in great detail. Composite structures are used on a lot of airliners. We have to know what happened to cause this accident.