is that the lift acts through the CP to create a torque about the CM
in the opposite direction to the torque that initially caused the
rotation of the rocket.
Not really lift (well it us, but), it's simpler.
Imagine a wind that blows at right angles to the fuselage. The magnitude of the force it causes is a function of the wind speed, the cross-sectional area and the fuselage shape (streamlining). For a first-order approximation, you can ignore the shaping, and assume the wind is less than the airspeed of the rocket, so the force becomes a function of cross-section. If you then integrate that force over the entire rocket, you'll find there's more force where the rocket is wider (the fins) and less where it's not (the fuselage and nose) and the net result is uneven. This can be reduced to two forces, a net sideways force and a torque.
When the rocket is aligned with the relative wind, then, as seen from the air's point of view, the rocket is completely symmetrical -its a circle with some very thin fins sticking out evenly around it. It's only when there's a sideways component -wind- that the "apparent shape" becomes non-symmetrical and there's a torque. And that only lasts until the rocket rotates into the apparent wind and becomes symmetrical to it.
but it can't be large enough that it causes significant
overcompensation (causing the rocket to pass vertical by its
Rockets generally have very little rotation momentum compared to the force of their fins. It's an issue to consider, but generally speaking, if it's stable in the wind, it's stable.
I do need to point out that such a case is not, as you put it, because it's too stable. The fins always work to align the rocket with the wind, putting on too much fin will lower performance through drag and trajectory issues, but it won't make it less stable.
Also, what sort of torque values are likely to be induced to the
rocket from wind and engine/thrust inconsistencies
In your example, this is much more important. When you have multiple engines in a cluster you have to consider stability in an engine-out condition. This might happen right off the rail if one of the igniters fails or simply takes too long, but it also happens at the end of firing if one engine is running a little hotter than the others.
For stability, you can consider the worst-case scenario. So if your engines are arranged in a triangle, for instance, you want to know what happens when one of the engines fails. This is simple vector decomposition - you have 2 x F(r) on one side of the triangle and 0 on the other, acting through the center of mass some distance forward of the engines. This will tell you the torque from the asymmetry.
Now the problem is calculating the torque from the fins. This varies widely based on the design of the fins. NARCON has some stuff on this, but it gets complex really fast. I recommend reading the section on the topic found here, although this is J class engines, not B's.