Electric dipole transitions: why is $\langle \vec{d} \rangle$ for stationary states 0 but not for superpositions?

Question

We've learned in lecture about electric dipole transitions for hydrogenic atoms, and our professor claimed that $\langle\vec{d}\rangle=0$ for stationary states, but not for superpositions of states. This makes entire sense for stationary states: For a stationary state, $\langle\vec{r_e}\rangle=0$ because the electron in the hydrogenic atom has no spatial preference and thus has an expectation value of 0. It then follows that: $$\langle\vec{d}\rangle=\int_{\mathbb{R}^3}d^3\vec{r}\psi^*\vec{d}\psi$$ $$\langle\vec{d}\rangle=\int_{\mathbb{R}^3}d^3\vec{r}\psi^*(q\vec{r_e})\psi$$ $$\langle\vec{d}\rangle=q\int_{\mathbb{R}^3}d^3\vec{r}\psi^*\vec{r_e}\psi$$ $$\langle\vec{d}\rangle=q\langle\vec{r_e}\rangle=q(0)=0$$ But why doesn't this same logic apply to superpositions of states? A mathematical answer would be nice, but I'd REALLY appreciate a more conceptual understanding -- what is it about superpositions of states that allows the electric dipole moment to have a spatial preference?