... What's more, because the horizontal force is
acting through a point behind the CG, the plane would spin in the
opposite direction to the force.

And when the plane is flying straight and level, since the wing is
behind the CG, shouldn't the wing, similarly, be continually pulling
the tail up over the nose, leading the plane to tumble through space?
Why doesn't it? Because while the wing is aft of the CG, the tail
provides a downward force to balance it. Since the tail is so much
further away from the CG than the wing is, it can leverage a relatively
smaller force to balance out the torque, while allowing the much larger
force at the wing to provide a net upwards force to support the plane's
weight.

When the plane is in a bank, yes, the wing is providing a horizontal
force from behind the CG again. But nothing has changed from the
straight and level case, there's still an upside-down wing back there
at the tail, and it's still balancing out the torque produced by the
wing being behind the CG. When the wing is banked, the tail is banked
too.

It's not easy to imagine this, but think of the space shuttle
stationary in space, then having a sideways horizontal force applied
behind the CG

But imagine another sideways horizontal force applied in the opposite
direction at the tail, exactly strong enough to balance the "torque".
If your left force is applied aft of the CG, but fairly close to it
(like where the wings are), and your counter-balancing right force is
applied at the tail, much further from the CG, then a relatively small
force at the tail will be enough to counter the much larger force at
the wing, due to leverage.

So our torques are equal, preventing any rotation, but there's still a
net force to the left (the force up near the wing is much larger), so
there will be a net translational force to the left. Your space ship
will accellerate to the left without rotating.
-harry