The researchers looked at how differences in the distribution of p-values can affect a test called the Kolmogorov-Smirnov test. They found a way to set a limit on how much the test statistic can vary when the null hypothesis distribution is not exact. They also suggested using a different version of the test to improve accuracy.