What changes could be implemented to fix the threats to validity

What changes could be implemented to fix the threats to validity (i.e. randomisation, history, experimenter affect, ?,(any problems that might be an issue with making these changes) eg. ethical problems?