I don't know of anything similar for Jupyter notebooks or sql either. I guess testing with simulated data designed to yield expected results would be the way I test my stuff.
Exactly. It feels like we're falling into the exact same problem with Jupyter notebooks as with spreadsheets: they become increasingly used by professionals who code (but who aren't software devs) to create bug-ridden, unmaintainable, large-scale software because they become familiar with the tool they have at hand.
I'm hesitant to blame the tool. Instead, I think it's a matter of our exuberance and interest in producing new results that causes us to get ahead of our software engineering skills and build things that get out of hand. Also, the professional developers are simply not available to help us improve things. We're on our own.
The really cool thing here is that they parse the excel formula in order to automatically “figure out” how they can perform the bootstrapping.