All these theories about auto pitch trim behaviour. I just think that they fixated on the ASI and pushed and trimmed until the speed started to build up. At that point, the pitch attitude was probably too low and the pitch trim too far forward for recovery at that altitude.
Remember how they got into this situation in the first place: On the go-around, they let the nose go up to 25 degrees until the speed decayed down to 125kts. I think that means that they would not have paid any attention to the ADI. Neither then, nor at the summit when the aircraft went into negative pitch.
Almost all the captain's 2500 hour flying experience was on the 737, and less than 500 hours in command on it. With the ease autoflight makes flying these days, and company policies that discourage hand flying, they couldn't have had that much experience at hands on raw data instrument flying.
Worldwide, the number of accidents due to loss of control because of poor instrument flying skills is becoming a concern. Automation makes things easy, but losing it makes it far worse than never having had the luxury of having it in the first place. I can't imagine a raw data instrument scan to be up to speed when the only hands on flying done is the take off up to 200-400 feet, and landing below 1000 feet whilst following the flight director.