While you're right, most people on here try to solve issues as consice, clever or creative :)

So I agree, it's not good practice, and one should never use code like this in production.

I don't think this solution is good practice. It's clever. It's efficient (if you need it). But in most projects it's just trick programming that makes code harder to read.

It's not the tests that are broken, it's your solution. Maybe you misunderstood the task.

I copied your code and ran it, here's the first bit of feedback:

Incorrect answer for n=195: expected 15 to equal 6

Your returned 15 is more than 1 digit long. You need to repeat this process until you get one-digit number (15 -> 1 + 5 -> 6)

This kata is broken. The test results don't match the tests. I was able to make the tests work by updating the results in assert, but in the "attempt" it failed because again, results don't match the test. Example: "expecte 2 + 8 + 8 + 4 + 0 + 0 to be 6" - wrong test data and results.

Frustrating, as it was the first kata in js.