Purity conjecture for the Potts measure on random trees

Let $q$ be the number of Potts symbols, let $\theta=\tanh(\beta)$, and let $d_0$ denote the random offspring degree under the tree measure $Q$. Reconstruction is the problem of recovering the root state from the states arbitrarily far from the root. Purity conjecture. For $q\leq 4$, there is $Q$-almost surely no reconstruction if

$$\mathbb{Q}(d_0)\left(\frac{2\theta}{q-(q-2)\theta}\right)^2<1.$$

This conjecture asserts sharpness of the Kesten–Stigum bound in the stated regime for random trees, extending the corresponding conjecture of Mézard and Montanari for regular trees. The paper notes that numerical evidence supports the claim, while results cited there establish sharpness for $q\leq 3$ only in sufficiently large degree and non-sharpness for $q\geq 5$; the stated random-tree claim remains open in the source.