I looked through the test cases and I can't see how the combinations you list could appear.

I guess you found them during the first batch of random test cases, but there the first parameter to is-merge is always either "Can we merge it? Yes, we can!" or "Can we merge it? No, we can't".

I even ran through the test cases 10 times to see if I could recreate the problem, but it seems to work fine for me.

Could it have been a temporary hiccup on the servers?

Please try again and let me know if you still see the same problem.

It does look strange, indeed. I'll try to look into it later today.

Thanks for pointing it out. :-)

Maybe you are calculating a bunch of Fibonacci's numbers?

Nah, he's calculating the number of kittens. Seriously, some of your comments are completely off :D. But for those, who're interested, g964 means "Do you memoize the whole Fibonacci tree? That takes too much memory. There's an approach that takes O(log n) operations to get the nth Fibonacci number, which should help you to finish in time."

Maybe you are calculating a bunch of Fibonacci's numbers?

If so, do the non-JVM versions of this kata have similar restrictions?

Kind of. There's definitely a limit in Haskell, Python and Javascript, although I'm not certain of the actual numbers.

I don't see why you got "expected fractions" with denominator 136240, the "expected result" is with a denominator of 34060.
Aren't you confused between "expected" and "actual"?
Read the post from jacobono below. Seems you have lots of problems with my kata these days!:-)

Please, have a look at the drawing. Your algorythm must follow the drawing, the description says:

Can you translate this drawing into an algorithm?

Your algorythm is certainly good but it doesn't answer the description:-)
PS: it's not an issue but rather a "Question"...