Once this happens, the autopilot, sensing a nose down trend, reacts by bringing the nose up again to level attitiude. This cycle repeast over and over throught the approach and this is all done without the pilots even noticing because it's so subtle.

Since the autopilot has brought the nose up again, the airspeed has decreased a few knots (or more) each time and you (the pilot) hasn't added any power to increase airspeed.

All this time the autopilot has been diligently trying to counter the decreasing lift by pulling the nose higher and higher resulting in airspeed bleeding off to just a few knots above stall speed.

I am more curious about what happens when the autopilot disengages on its own as it appears to have done in this case.

Would this leave the aircraft in an out of trim position requiring the pilot to apply considerable inputs that might be surprising to a pilot?

Through about 8000 feet, the airspeed started a gradual decrease from 170 KIAS over a period of five minutes. During this time, the vertical speed continued at a constant 1190 fpm up. The gradual decrease in airspeed was detected when the first officer looked up from his paperwork, noted the decreased airspeed, and advised the captain. The captain then rotated the pitch control wheel on the flight guidance controller toward nose down (see Figure 1) to increase the airspeed. While attempting the adjustment, the captain saw the aircraft's stick shaker activate, causing the autopilot to disengage. This occurred at 14 800 feet above sea level, at 104 KIAS. The captain then began to manually fly the aircraft.

Within a second of autopilot disengagement, the aircraft began to roll right and pitch down (see Appendix A - Flight Data Recorder Plot and Appendix B - Flight Data Recorder Plot - Engines). Immediately after the aircraft began to roll, it was noticed that there was ice on the left engine inlet. The roll angle increased to 64º, the pitch angle went from 15º nose up to 5º nose up, and the aircraft vertical acceleration dropped to approximately 0.5 g. The aircraft pitch then increased to 30º nose up briefly before decreasing to 40º nose down. These conditions are indications that the aircraft wing had fully stalled. However, the captain interpreted the indications as severe turbulence.

The FDR data show that the aircraft underwent three distinct stalls during the loss-of-control event, with the third stall being the most severe. The data show that the control column position cycled rapidly back and forth as the stall developed, but was moved generally aft, remaining aft during all three stalls. There was significant aileron and rudder pedal movement during the event, but these controls were ineffective in regaining control and were in response to the aircraft's movement, rather than the cause of it. The data indicated that aircraft control was regained when the control column was moved forward.

...

Immediately after recovery, the crew observed that ice was building up rapidly on the aircraft fuselage. Airframe de-ice equipment was then selected ON. The pneumatic boots functioned when selected, and were effective in removing the ice. The crew requested a lower altitude to remain clear of icing conditions and continued to Deer Lake. After landing, the pilot reported a severe turbulence encounter to company personnel. A heavy turbulence check was subsequently carried out, and no damage was found.