The previous version was not correctly assigning the errors from single value validation variables. Instead they were always from previous run and not the actual run no.