Rotation of rigidbody composed of point masses - weird behaviour

Hey,

I have a set of loose point masses. I integrate velocities like this:

for (int i = 0; i < pointMasses.Length; i++)
	pointMasses[i].velocity += pointMasses[i].force * dt / pointMasses[i].mass;

I thought about making an experiment. I took 4 points masses and positioned them such that they compose a box. I wanted to use rigidbody formulas to make the “rigidbody” rotate correctly, i.e. determine point masses's points such that after integration they still maintain distances between each other unchanged + angles.

The basic formula for velocity of a moving point (Q) relative to rigidbody's center …