At last I have in my quest to find an official definition of "upset" it is here in the following document which is worth careful reading.
https://www.faa.gov/other_visit/avia...overy_book.pdf
An airplane upset is defined as an airplane in flight unintentionally exceeding the parameters normally experienced in line operations or training. In other words, the airplane is not doing what it was commanded to do and is approaching unsafe parameters.
A few examples amongst many in this document:
In another high altitude situation, the crew decided to use heading select mode to avoid weather while experiencing turbulence. The steep bank angle that resulted from this mode quickly caused slow speed buffeting. The crew’s rapid inappropriate response to disconnect the autopilot and over-control the airplane into a rapid descent in poor weather exacerbated the situation. These real world examples provide evidence towards the need for more detailed training in high altitude operations.
For example, a recent incident occurred where an airplane experienced an environmental situation where airspeed slowly decayed at altitude. The crew only selected maximum cruise thrust, instead of maximum available thrust, and that did not arrest the slowdown. The crew decided to descend but delayed to get ATC clearance. Airplane slow speed buffet started, the crew selected an inappropriate automation mode, the throttles were inadvertently reduced to idle, and the situation decayed into a large uncontrolled altitude loss. This incident may easily have been prevented had the flight crew acted with knowledge of information and techniques as contained in this supplement.
and this most extraordinary example:
An airplane was approaching an airfield and appeared to break off to the right for a left downwind to the opposite runway. On downwind at approximately 1500 ft, the airplane pitched up to nearly 60 deg and climbed to an altitude of nearly 4500 ft, with the airspeed deteriorating to almost 0 kn. The airplane then tail-slid, pitched down, and seemingly recovered. However, it continued into another steep pitch up of 70 deg. This time as it tail-slid, it fell off toward the right wing. As it pitched down and descended again, seemingly recovering, the airplane impacted the ground in a flat pitch, slightly right wing down. The digital flight data recorder indicated that the stabilizer trim was more than 13 units nose up. The flight crew had discussed a trim problem during the descent but made no move to cut out the electric trim or to manually trim. The accident was survivable if the pilot had responded properly.