You can prepare a pair of photons in a state such that when you measure the polarization of both of them along the same axis, for whatever direction you want to choose, you get the same result. But they are entangled, each photon considered separately is not in a well-defined state.

You can also prepare two photons in the same state, so the have the same polarization for some direction chosen at that time. But the measurements along other axis won't be perfectly correlated (if they are correlated at all).

The red/blue color example is too simple to be interesting.

You cannot paint a billiard ball with a superposition of colors.

But what is the implication of that? What do those words cause to be different in practice?

The randomness in the observed color of the ball. (The observable represented by the "color operator" could have two or more distinct eigenvalues.)