The TE flaps shown in the photos of the wreckage are not extended; the LE devices (Slats outboard, Krueger flaps inboard) are extended.
The TE flaps are driven by screw jacks turned by a hydraulic motor. They will either remain in the position they were driven too or break off completely. An impact cannot knock them back to the stowed position.
Conclusion. Either a) The flaps were not extended before landing or b) they were retracted by the crew during the landing roll in preparation for a go-around.
A credible witness reports that he noticed the flaps remained stowed during the approach. This confirms that option a) is the case.
There are several reasons for landing without deploying the flaps, but we have some indication that the crew experienced a flap problem prior to landing but "had dealt with it" - using the QRH no doubt. Since neither the primary nor alternate flap drive systems deployed the flaps, one is led to suspect that the problem was an asymmetry lock-out. This is most commonly a sensing system defect - genuine asymmetric deployment is extremely rare. Flap position indicator faults are not - the weak point in the asymmetry sensing system has always been the flap position indicator.
For an asymmetry sensing fault, the flaps would have had to move at least a small amount. Therefore, another factor that I feel supports this hypothesis is that although the flaps shown in the photos are not extended, neither do they appear to be fully retracted.
Once committed to a flapless landing, the pilots must be very careful about controlling speed during the approach. For obvious reasons practice flapless landings are usually conducted in a simulator. Is the actual handling of a B737-400 making a flapless landing matched by the simulator motion systems?
I'm a Technical Services Engineer with FDR readout experience and trained in accident investigation, not a pilot. Perhaps a B737-400 rated pilot can comment on the simulator/aircraft handling comparison?