Why is sound absorption by tunneling defects the same in all glasses?

Below $\sim 1\,\mathrm{K}$ amorphous solids show specific heat $C \sim T$, thermal conductivity $\kappa \sim T^{2}$ and an internal-friction plateau $Q^{-1} = (\pi/2) C_{\mathrm{tun}}$, with tunneling strength $C_{\mathrm{tun}} = P \gamma^{2}/(\rho v^{2})$ between $\sim 1e-4$ and $1e-3$ across chemically different glasses ($P$ = density of states of tunneling two-level systems, $\gamma$ = deformation potential, $\rho$ = mass density, $v$ = sound velocity); equivalently the phonon mean free path is $\sim 150$ wavelengths. Explain this universality, e.g. by interaction-induced renormalization of the defects (Yu and Leggett; Burin and Kagan) or by a generic soft-mode mechanism, and predict where it fails.

A theory deriving $C_{\mathrm{tun}} \sim 1e-4 \text{ to } 1e-3$ from generic properties of glasses, confirmed by a predicted exception or by simulations that compute $P$ and $\gamma$ in several model glasses.