If the depicted maneuver did happen it probably happened in quite a slow speed and such a quick recovery was a fluke, not something you can easily reproduce, or even want to.
I'd agree to that upto a certain point.
Among the two depictions Cpt. Bryan, in his last-ditch attempt in a wild snaproll, lost control of the plane and fell straight downwards. Widewing, in one of his films, also loses control of his plane after the desired motion, and plummets for a while until he regains control. In case of Lt. Candelaria, I sincerely doubt he predicted the turn of events as he went into the maneuver, and my reasoning is his hybrid form of a barrel roll/snap roll was done with wide arc/radius as the enemy overshot, which landed him a one in a million shot.
In any case, (IMO) there's a reason why "snap roll" type of maneuvers never really makes it among the list of "official" ACM, since it is a result of a accelerated stall, rather than a plane flying within the the limits of its envelope. It is because the plane steps out of the envelope, that it rolls so fast, way faster than its normal rolling speed, in the first place. And as bozon once said, and I quote, "anything can happen when normal airflow is lost over the plane".
However, how a plane reacts under such conditions, whatever dangers there may be, is not entirely random. In some cases it is predictable, and "snap rolling" is a technique used by many pilots both in the game, and in real life. Would it be so hard to believe if it was just a normal snap rolling? People do it all the time.
Something not easily reproduced, I'll agree to that. However, it's not as if you are deliberately crashing your plane into a mountain. In both cases, Cpt. Bryan and Lt. Candelaria mentions they've practiced it before. How'd they practice such if it was something unreproduceable in the first place?