The set that you are using is hash set, operations of which does not work in $$$O(\log n)$$$. Hash set has operations that in theory work in constant time, but in some testcase, they might work in linear time. The testdata might contains such test, and they often do because the problem setters oftern make anti-hash test.

For $$$O(\log n)$$$ set, you might want to use TreeSet.

It's actually not the issue of the set but actually the issue of sorting. Before Java 14, Arrays.sort on primitives gives worst case quadratic time. Shuffling before sorting gives AC 110982385. On a side note I'd like to point out that Java HashSet is actually only worst case O(logn) performance on comparable elements (e.g. Integers). This is because it becomes a BBST upon getting overloaded.

I also got hacked once because of using Arrays.sort() to sort a combination of numbers.

Since then, I do the following two things interchangeably.

1. Using Collections.sort() by using data structure as ArrayList.
2. First shuffling the array and then using Arrays.sort() to nullify the chances of hitting the worst case scenario in double pivot quick sort.

    for (int i = 0; i <= a.length - 2; i++) {
       int j = i + r.nextInt(a.length - i);

       swap(a, i, j);
    }
}

public static void swap(int[] a, int i, int j) {
    int temp = a[i];
    a[i] = a[j];
    a[j] = temp;
}

Credits: This comment