Thanks for the reply, and after careful rewatch, I agree with you on some of the points, but not the major ones. My background comes from a past life of years teaching acro and unusual attitude recoveries in light pistons through light jets. (These days, I accept flight plan datalinks into the FMS and click TOGA for a living.) And my biggest lasting impression is the difficulty of breaking people away from their gut response to instinctively pull on the elevator any time stress and/or unusual situations are encountered, ranging from acro upsets to normal flight low speed events.

First points of agreement are that

- this (initial) upset is very expected from stalling at high sideslip
- their correct initial response to center everything, along with pushing slightly to reduce AOA, and avoiding aileron until clearly out of the stall
- and then to pull the thrust to idle after they've taken a breath and evaluated the face full of ground in front

(However you note that they then pulled slightly (0:13) to avoid excessive acceleration, which, being inverted, would pitch them into the ground and increase acceleration. I think it's likelier they were just pitching to stop the visual attitude rate and get a stable platform from which to continue.)

After that comes the first biggie, the roll direction: at "easy roll to the horizon" (0:15) the nearest horizon is clearly to the left, and the airplane rolled *away* from that to the right, toward inverted. And unlike what I remember from past watching, and unlike what you said, this didn't seem to be from yoke input at this low speed, which stayed centered (plus and minus some transients.) So it's probably from residual sideslip.

Next, the pull. Here, I partially retract that they did a rolling pullout. The yoke is centered, and from the PFD it's hard to tell. It shows a right roll, but from a clean pull, it could still falsely show roll as they pass through the bottom near (but not exactly through) the nadir. (The animation to the right is worthless for this without knowing the camera origin point and other geometry factors.) I was prepared to fully retract my statement, if it wasn't for them seemingly showing intent to roll with the hard to make out "easy right... easy right turn there." 

Stepping away from this event, as a generality we know that for nose low inverted upsets, we need to roll upright and then pull out. But also, if the pitch attitude is close to vertical, let's say 70 degrees nose down inverted, it would be quicker in time (and in less airspeed built) to pitch 40 degrees through the bottom to 70 degrees nose down upright, than it would be to roll 180 degrees to the same attitude. And this is indeed the typical scenario (and recovery) from a half turn spin in a GA trainer.

Back to this video, at initial stabilization they were at a moderate 20 degrees pitch down, from which I would have absolutely rolled in any circumstance. But not having done it in an airliner, I entertained the possible explanation that there were particular factors I hadn't considered, and that they had planned and intended to pitch through in such a scenario all along. OK. But this is squarely contradicted by "easy roll to the horizon" and "easy right turn there," so I'm back to no explanation, and defaulting to my original explanation: that they planned to roll (the conventional response), were a few seconds behind in processing and began this roll after the airplane had already rolled itself (from residual beta) through inverted and was quickly pitching itself (from the trim) through 45 degrees, and at the point they began the roll the nearest horizon was to the right, toward which they rolled with a timid and invisibly small input (which, at these speeds, would be effective) simultaneous with pitching.

I hate to come off like I'm tarnishing other pilots' actions. I recognize that I have the benefit of watching this from a chair, and can only hope that my hundreds or thousands of times of doing this planned would result in a correct response if I ever encounter it unplanned. But in the end physical facts are physical facts, and if someone asks me about this video with respect to what they should do in that situation, I can only point to it as an example of what to avoid.