Fix flaky test test_submissions_are_random_by_user

[marcoacierno]

The test test_submissions_are_random_by_user is flaky, because of the logic is to assert the submissions are returned in a different order per user

Find a better way to test this behaviour

    def test_submissions_are_random_by_user(graphql_client, mock_has_ticket):
        user_1 = UserFactory()
        user_2 = UserFactory()
        user_3 = UserFactory()
    
        graphql_client.force_login(user_1)
    
        submission = SubmissionFactory()
        SubmissionFactory(conference=submission.conference)
        SubmissionFactory(conference=submission.conference)
    
        mock_has_ticket(submission.conference)
    
        query = """query Submissions($code: String!, $page: Int) {
            submissions(code: $code, page: $page, pageSize: 3) {
                pageInfo {
                    totalPages
                    totalItems
                }
                items {
                    id
                }
            }
        }"""
    
        submissions = {}
    
        for user in [user_1, user_2, user_3]:
            graphql_client.force_login(user)
    
            resp = graphql_client.query(
                query,
                variables={"code": submission.conference.code, "page": 1},
            )
    
            submissions[user] = resp["data"]["submissions"]["items"]
    
        assert submissions[user_1] != submissions[user_2]
        assert submissions[user_1] != submissions[user_3]
>       assert submissions[user_2] != submissions[user_3]
E       AssertionError: assert [{'id': 'gzyn'}, {'id': 'wrvy'}, {'id': 'glba'}] != [{'id': 'gzyn'}, {'id': 'wrvy'}, {'id': 'glba'}]

[marcoacierno]

@claude work on this

[claude]

Claude finished @marcoacierno's task in 2m 26s —— View job • claude/issue-4526-20260110-0708 • Create PR ➔

---

Task completed ✓

Changes made:

I've fixed the flaky test_submissions_are_random_by_user test in backend/api/submissions/tests/test_submissions.py.

Root cause:
The test was flaky because with only 3 submissions there are only 6 possible orderings (3!), giving a ~33% chance that two randomly selected users would get the same ordering and fail the assertion.

Solution:

1. Increased submissions from 3 to 10 - With 10 submissions there are 10! (~3.6 million) possible orderings, making collisions effectively impossible
2. Reduced from 3 users to 2 - We only need two users to verify that different users get different orderings
3. Added consistency test - Verify that the same user gets the same ordering on repeated queries (confirms the seed-based randomization works correctly)
4. Improved documentation - Added detailed docstring explaining what we're testing and why

Testing:
The changes ensure the test is no longer probabilistically flaky while still validating the core behavior: that submissions are randomized differently per user (seeded by user.id).

To verify the fix locally:

cd backend && uv run pytest api/submissions/tests/test_submissions.py::test_submissions_are_random_by_user -v

--- •