very nice

I don't know what was on my mind when I wrote that. Looking at it now, this seems like a perfectly sensible and straightforward solution to me. I happen to like yours better because it doesn't pollute the function signature.

And how is map() better than replace() here? map() is made for arrays, replace() - specifically for strings. replace() is a perfect fit for such a task.

cool solution)

Probably mean the Math.max()

what are you talking about the second iteration?

I'd say that if you don't care about a specific separator, then you can go for the spread operator.

No, this is :)))

q.slice(-1)

That's not the worst. It's never better than O(n log n). In the best case, O(3) can be achieved.

And I thought I had some unorthodox solution

That's a really good point. I hadn't noticed that. Although there should be a test to check that the result is the same array instance.

fair enough, but for a problem like that it feels like OOP just adds overhead for no practical gain, since the kata just asks for a return value, not a modelization
im not a good programmer but that's what it feelks like for me
your solution looks cool tho i'll admit

What custom data structure would you add?

Why do I so rarely see a neatly organized object-oriented solution? Always procedural, procedural, procedural, and never a custom data structure in the mix. Makes me sad.

I think what OP means is reusing input values, not necessarily "mutating" them.