I did the same. After reading description I understood it just like this: there should be one step undo and (therfore) one step redo. Maybe that would be good to address this in the description somehow?

I have added 4 tough cases that change false to true or vice versa if you add/remove just one balloon. Now your 85Kg solution won't pass. Thanks for pointing out!

The idea was to let people implement their solution in either traditional function-based way (via function based class) or with new ES6 classes. That actually what I observe in solutions: some people create functions and some classes. I've seen many katas operating the same words (asking for class and suggesting functional implementation) like this one https://www.codewars.com/kata/object-oriented-piracy. Can you advise how can I keep this flexibility while avoiding possible confusion?

Wow, thank you! I have to build several really precise tests. All of the existing tests are indeed too loose.

Great solution!

Just a minor objection as just a few hours ago I watched Douglas Crockford talking about potential bugs that can emerge from using ==. He suggests using === instead. Always.