An alternative view of flight control in conventional aircraft is that stick position via cables / rods moves the elevator / tab. Force is a necessary feedback for structural limiting (g) and speed stability (offset trim position). There are many compensating mechanisms - temp / cable stretch, springs for neutral position / stability, alt-speed for structure (q feel).
Offsetting the pitch trim in level, constant speed flight will change the stick force, but does not change the stick position (depends on aircraft type, also ignoring inevitable ‘small’ changes due to mechanisation - trim / elevator).
The aircraft flightpath does not deviate. It can be controlled quite adequately, but requires a pilot compensating force for a given stick position - having to fly out-of-trim.
Without force compensation the aircraft will deviate because the corresponding stick position - position of zero force has changed.
With a trim malfunction pilots are required to apply a compensating force for an ‘apparent’ correct stick position; but the aircraft has a different, and on occasion variable stick position / force relationship.
This requires considerable mental compensation - change from the normal relationship, and applying ‘unnatural’ force / direction. The perception is that aircraft is not controllable according to previous experience and training, and by reversion to the norm the pilot inadvertently seeks to fly zero stick force, allowing the malfunctioning trim to fly the aircraft.
Auto thrust / speed can further complicate the pitch control relationship.
This is a difficult dynamic interaction to describe.
Simulator flight with an offset trim provides a good representation of the revised piloting task (but without surprise), similar to flight with trim failed fixed.
Repeating the demonstration with varying trim position (737 MCAS not yet in simulators?); then the piloting task is to generate a new model of how the aircraft should be controlled for each errant trim setting. In addition the stick force / datum position will change with speed, which also changes because of inadequate aircraft control, which also changes with pilot trim input, and further errant trim input, and … , …
There is no specific anchoring point for reference.
An alleviation is to reference a constant attitude, move the stick to control the aircraft, ignore any force reference. This requires considerable mental effort, which could be already challenged by the circumstances leading to the trim failure (is the aircraft really changing state or just the instrument indications - AoA error generating an UAS like situation, MCAS trim only active with flap up).
Thereafter determine the source of apparent control problem, and if trim related, isolate it.
The speed (altitude, config) will generally determine the maximum trim force which can be held and still manoeuvre the aircraft - depends on specific aircraft design / configuration. If excessive then certification would require an alternative trim mechanism.