Vessbot, thanks for that view.
Semantics aside, the speed based deffiniton is widely used thus might have helped with a search; apparently not so.
The AoA aspect, not debated, also reflects differences in viewpoint. Theory or practice, steady state vs the dynamics of real situations which crews have to manage.
Limit points are usually static - steady state; thus in flight tests the ‘crossover point’ can be defined with smooth coordinated cross-control inputs without changing bank angle or speed (AoA). However, in reality if an aircraft is suddenly yawed - rudder failure, the dynamics of yaw to roll, speed / AoA change, as you note, could overcome the roll control power. Thus although an aircraft is within a limit condition, flight control is lost because of the combined need to stop the roll and recover to wings level.
This supports the comment by The Rt Hon minister (#11), that simulators, particularly training devices are no more than best guess; even more so if the manufacturer has not established the limit condition or recorded the dynamics of off-limit conditions.