In the early days of the Avro RJ, the autoflight system was modified after an interesting (serious) event involving a 3000ft+ pitch up and a couple of impressive wingovers. This was due to an inadvertent autopilot engagement and subsequent overpower – back driving the autotrim nose-up to a point where full forward stick (position and force), could not maintain control. Manual trim did not ‘appear’ to work because unbeknown the autopilot was engaged.
IIRC the autopilot automatic disengage logic now contains displacement/rate/duration logic (trim and servo torque) to prevent hazardous overpowering, but this is not intended to be an emergency disengage function. Disengagement is via one of two stick top buttons (dual paths), or an emergency switch on the control panel.
Again during the aircraft’s history the AP disengage buttons were modified to reduce inadvertent disengagement (white knuckle syndrome), by fitting a higher guard or lower button, not sure which. However, for some people this made disengagement more difficult (fat thumb syndrome) and thus the necessity to remind pilots to look (and hear) confirming indications of AP disengagement.
AFAIK the failure of a AP servo to disengage would be annunciated with a permanent red AP light or ‘normal’ indications that the AP was still engaged, in either case the panel switch should be used.
If the autos have disengaged, then a ‘conventional’ control-system jam might be suspected. The pitch-control system can be split by overpowering with the other stick or using one of two emergency disconnect handles.
The elevators are servo-tab controlled and are free floating about the fixed horizontal tail; thus, a stick jam requires a restriction in the stick/cable run (autopilot servos-motors are in one cable run).
If both servo-tabs had jammed, the stick / elevator should still move, but with very high stick forces - normal aircraft motion. An autopilot servo-motor ‘jam’ would be like a single servo-tab jam as the AP only drives one elevator servo-tab; limited normal aircraft motion (and roll from the single elevator).
For both of the (independent) elevators to freeze requires a significant icing event (or rags around the elevator hinge – it happened). In this circumstance, the stick should still move and displace the servo tab, and even in the extreme of both servo-tabs freezing as well, then the stick should have some movement against the servo springs, but without aircraft motion.
There is an independent elevator trim system (trim tab) directly to the elevators.
Stick jam – elevator free, then the trim moves the aircraft in a conventional sense. Stick jam and both elevators jam, then trim has a small effect on aircraft motion but in a reversed sense.
(It’s late in the day and with tiring mind, I reserve the right to revise/reverse the above).
The RJ has a highly redundant manual control system; failures require a bit of force to ‘break-out’ and split the system, but thereafter fly whatever remains active.