Open problems in applying random-matrix theory to nuclear reactions