It looks like the protection will trip if the average of the two outputs is much away from 0V.
Both speaker output terminals should read 0V, or close to that; when there is any audio out they should swing opposite polarities, so the average stays around zero.
If one is stuck with a voltage on, with no speaker connected, start working back through the power devices and the two driver transistors for each of those, looking for shorted ones?
Or try disconnecting both output transistors on the dead side and check for AC between source and gate connections of each - it's class D, so I'd expect a continuous signal for each device?