I don't know how to test this (in the linked web site, I don't see an execute button, nor a place to put the arguments), but from the description, I doubt that it gets the rounding right. Does 2/3 really give the output from Example 2, not one bar less?

I also couldn't see a way to run the code at the time the answer was posted, but there seem to have been some improvements to the interface since then. There is now a "Run" button to the lower right of the code, which displays the output for the 4 test cases.

One of the test cases is off by one, so I've added a separate thread (so that it is immediately clear from the thread title that something definitely needs to be fixed).

The challenge author has now edited to allow rounding of 33.33 to either 33 or 34, so there is now nothing to be fixed. This answer is correct on all test cases.

I don't like the challenge terms to be changed after examples have been posted (except to clarify the task if the original text was unclear). Even if the change doesn't invalidate examples, it penalises those who took care to implement the previous specification correctly. Definitely in terms of time spent writing the solution, but potentially also in terms of score if the new specification allows a shorter implementation.

From my perspective, the new wording appears to be a change, which I'm uncomfortable with for the same reasons as you.

However, my impression is that from the challenge author's perspective, this is a clarification rather than a change. The wording in the comment on the challenge supports this:

The rule means that 33.x can be rounded either down to 33 or up to 34, no matter what digit x is. I can add second test cases if needed.

It's unfortunate that this was not caught at sandbox stage. I didn't suggest an additional test case output at that point because I was not yet aware of the ambiguity. I thought the single output for that test case was the only valid one. To me, "rounding" refers to the treatment of the half way point only, but the existence of several answers rounding 33.3 up to 33.4 suggests not everyone uses the word in the same way.

In general, I would much prefer to avoid rule changes. In this particular case, I can't see a way to have spotted this ambiguity before answers arrived, and now that answers have arrived, either way of resolving the ambiguity will disadvantage half of the answers (it was just a matter of choosing which half).