still O(n^2)

unless I'm understanding the problem incorrectly

Exactly. This has been discussed hundreds of times already in the comments.

Javascript Test Case: Test.assertSimilar(dirReduc(["NORTH", "WEST", "SOUTH", "EAST"]), ["NORTH", "WEST", "SOUTH", "EAST"])

These directions should reduce to [], but it wants me to not do anything to it. This is surely an issue, unless I'm understanding the problem incorrectly.

nope. Sounds like the JS version is flawed (see opened issue below)

Is this supposed to happen?
Your warrior should now be LV 3 - Expected: 4, instead got: 3
Because if my warrior should be level 3, then it shouldn't have to be level 4 right?

With the performance requirement, I beg to differ. ( That requirement should be documented better though. )

Needs performance tag. Needs clarification that every second person is to be unsurvived ( do not rely on examples - specify such things ).

With the performance requirement, it might have a place next to Giacomo's kata that is bloody linked from the "similar kata" under the description, but the performance requirement has to be clear. ( adding explicit text to the description would be preferable to just a tag! )

As said elsewhere, you're overflowing native Numbers. You could rectify this by supporting Node 10.x ( which you really should anyway ) and using BigInts.

But without random tests, you're headed for retirement mighty quick. Did you read the documentation on creating your first kata?

giacomo's, iirc

Needs random tests.

One of the test case has expected result 5534023222112866000, which is past the safe integer limit and so is suspected to floating point errors.

Initial code has the wrong function name.

Duplicate of what kata? ;-)

Duplicate.