Predicting universal transport coefficients in far-from-equilibrium quantum systems remains a fundamental challenge. A paradigmatic example is the non-thermal fixed point (NTFP) of isolated Bose gases, where coherence spreads as $\ell^2(t) = C\hbar t/m$ with a universal constant $C$. While the scaling exponent $z=2$ is well established, the amplitude $C$ has remained elusive because the underlying particle cascade $n(k)\sim k^{-4}$ leads to a divergent kinetic energy, threatening the very existence of a constant speed limit. Here we resolve this paradox and present the first analytical, parameter-free prediction of a universal amplitude $C$. A deep interplay between symmetry and dissipation is uncovered. The emergent weak U(1) symmetry at the NTFP enforces a conserved total current, forcing the low-energy phase dynamics to obey a diffusive Langevin equation with noise entering as the divergence of a stochastic current. This structure, combined with dynamical decoherence of high-momentum modes, yields a universal power-law momentum distribution $\tilde{f}(v)\sim(1+v^2)^{-3}$ (with $v=k\ell$) that naturally regularizes the ultraviolet divergence. From this, a parameter-free geometric baseline $C=3$ is obtained, independent of microscopic details. The experimental value $C=3.4(3)$ [Martirosyan et al., Nature 647, 608 (2025)] is then shown to be quantitatively consistent with universal logarithmic corrections arising from a marginally irrelevant coupling at the fixed point. A new paradigm is thus established for predicting transport coefficients in strongly correlated non-equilibrium systems: symmetry constraints determine the low-energy effective theory, dynamical decoherence provides a natural ultraviolet completion, and scaling analysis delivers testable predictions moving beyond scaling exponents to quantitative amplitude prediction. less