• Too few random tests
• Balanced/unbalanced random tests should be shuffled
• Returning multiple types (sometimes float,int, sometimes float,str) is bad practice

Kata should have random tests. See docs for reference
https://docs.codewars.com/languages/python/authoring/
https://github.com/codewars/codewars.com/wiki/Tutorial%3A-Create-Your-First-Kata#write-good-tests-and-utilize-both-fixed-and-random-test-cases

No random tests.

If function output has fixes quantity of values, it's preferrable to return a tuple instead if list, in Python.

2-sum is a duplicate to many katas.

test.assert_equals(get_addends(5, [2,3,0,5,9]), [2, 3])

Since when 2 + 3 != 5?

And this obviously contradicts the "No answer will have more than one two-number solution" statement.