This accident is strikingly similar to the Tatarstan Airlines Flight 363 B735 that crashed at Kazan, Russia in 2013. This is the graphic of the vertical profile in that accident:
Interesting, not seen that one before.
One possible reason for this profile is something I have seen in light aviation training - a confusion between stalling and zero g. If the push results in reduced g or slightly negative g, it can be percieved as the aircraft stalling, and so pic continues to push forwards to 'unstall' the aircraft. And the result is similar to that profile.
And this is not something that all pilots have trained for, as aerobatics are not on the syllabus. And is not something you can train for in the simulator. And there is a big difference in individual responses to odd attitudes and zero g. Some students shout 'yeeeehaaa' and ask to do it again, while others freeze. And it is not always the bold and confident student who shouts 'yeeehaa'. But if the reaction to zero-g has never been addressed and mitigated back in flight school, it is never going to be so in commercial flying. (Especially with spinning off the syllabus now.)
This is where a post-CPL course is aerobatics and gliding would be beneficial to all new candidates, to stretch their horizons and abilities. Experiencing an upset for the first time with pax on board is not the best time to discover you are uncomfortable with the sensation, or don't have an automatic stall identification and response reflex (Air France please note).