Imagine that if you'd have so much funding that you'd be able to cover a km2 in a forest with camera's and microphones, some remote controlled tools and enough food to experiment with. I think one could've made some kind of improvised tool that drops nuts when you handle a coin or something else. The only thing you then have to do is watch for a couple of interactions where one bird hands the other bird a bottle cap to exchange.

https://www.youtube.com/watch?v=LJG3282QU4g

Just extend the above experiment with such tracking and I'd think you can reach the same conclusion as this article had, but then with farrrr more comfortable ethics?