Uniform and subuniform posterior robustness: The sample size problem