Sorry I misunderstood the question first time; I thought you wanted to prevent rotation, not create some.
Well, at booster separation, Velcro Rockets applies a force for one timestep:
VECTOR3 pos, dir, rot;
GetAttachmentParams (attachpos[1], pos, dir, rot);
AddForce (_V(200*SepForce*GetMass()*rot.x, 200*SepForce*GetMass()*rot.y, 0), _V(0, 0, Tipoff*GetSize()/20));
SepForce is a scalar I set based on experiment, multiplied by the size of the sim timestep.
So, I can think of some ways to break this:
1. Use the Attach Point 0, for lower stages, not 1, for strap-ons (but you haven't done this).
2. Have an absurdly high Irot or RotResistance...but you haven't.
The only oddity I see is that the design intent was that the attachment points for the booster are supposed to be approximately at a physical attach point, and the point described in the scenario file is where that point should attaach to. In each case, the point should be on the outer skin of the cylinder.
But you have attached it at (0.05, 0, -5.0) on the booster. This is so close to zero that it effectively IS zero.
Try moving it to something like (1.5, 0, -5.0) and add the 1.45 to the attach points on the booster:
BEGIN_ATTACHMENT
C 0.0 0.0 13.96 0 0 1 0 1 0 VELCRO1
C 0.000 -1.183 -0.107 0 -1 0 0 0 1 VELCRO2
P 0.0 0.0 13.96 0 0 1 0 1 0 VELCRO3
P 0.000 -1.183 -0.107 0 -1 0 0 0 1 VELCRO4