At a height of 900 m there was a simultaneous control column nose down input and stabilizer nose down deflection from -2,5 deg (6,5 units) to +2,5 deg (1,5 units) (the FDR recorded a nose down stabilizer input from the stabilizer trim switch of the control wheel lasting 12 seconds, while the CVR record contains a specific noise of rotation of the trim wheels located on both sides of the central pedestal), as a result the aircraft, having climbed to about 1000 m, turned into descent with negative vertical acceleration of -1g.
I still think the answer lies in that quote.
If they pulled -1g to gain the level-off altitude,
I know what you mean, but 'pushed' might be more appropriate. But, moving on: I've read that the CVR recorded the captain's concern as to whether 8000 was set in MCP and the F/O's reply that "yes, here it is." If that was indeed the case why was there a control input at 900m to level off? The FD & HUD would still have been in TOGA climb to 8000. And for anyone to hold the pitch trim for 12 secs is unthinkable unless they were in a frozen state of panic. Both these guys had 1000's of hours and therefore I hope were competent & experienced.
I have been in a rapid climb, nose high, high speed ALT ACQ situation (B757) and reducing power was the game saver as the attitude was rapidly reduced, not trim.