Exploring the Gap between Tolerant and Non-tolerant Distribution Testing