How reliable is our existing ‘seismic hazard analysis'?