Absence of diffusion on the edge

The edge of a FQH droplet supports gapless excitations that are protected by a U(1) anomaly. At small but finite temperature, diffusive spreading is expected to occur around the chiral ballistic front. We show that this chiral diffusive fixed point is never stable. Hydrodynamic long-time tails give large corrections to dissipative transport on the edge, leading to a breakdown of diffusion. We comment on the connection to a possible lack of thermalization in the nu=5/2 state.