Trying to understand the rules for sorting two numbers with the same weight was really confusing because of the wording. I think this is an improvement:

When two numbers have the same "weight," sort them as strings and not numbers. For example, "180" and "90" both have the same weight (9), but when compared as strings "180" is first.

The issue isn't the number of tests, it's that they are random. You should do the same tests with the same input every time.

Non deterministic unit test are Bad(TM). Depending on the solution, the randomized tests may pass or fail from run to run. They should be changed to a cross product (testing all hands against all other hands).