On melt layer stability following a plasma disruption