Just because you're used to mutating the original input doesn't mean it's acceptable or it's a good practice.

You shouldn't be doing that at all, so take this as a lesson and learn how to write code that actually matters and doesn't bite people including yourself (comparing to, say, trivial issues like semicolons). Blaming the kata for your misconception is absurd.

If the kata description fails to mention key factors like this, would you consider the kata broken instead of the tests?
Because there IS a problem here. Just because people managed to solve it doesn't mean that everything is fine.

See the hint from "bkilrain":
"I guess they aren't passing because you are modifying the original objects passed in somehow. That wasn't specificied in the kata description but if you refactor to preserve the original objects, they will pass."

226 people (inclusive me before an hour) solved this kata without any problems.

So I think, the javascript random tests are not broken!