Under what circumstances can a very small treatment effect be statistically significant?
The ideal in sampling is to have as big a sample as possible and as small a variance as possible. The idea being that if you observe more and everything you observe is practically the same then you can be pretty darn sure of the population value.