Problem 174. Roll the Dice!

^Agreed.

Making a sophisticated test suite for such case(s) seems to be an interesting challenge!

@Dyuman Monte Carlo would probably work: call the function 1e4 times, check how often the two rolls are the same, and test whether the result is (statistically) significantly different from the expected value. Setting a very high significance level to avoid false positives would allow valid solutions to pass with a very high probability while catching ones such as this.