Why does sometimes a test case give the right answer when only provided as a single test case whereas fails to give the right answer when provided with multiple test cases?