Great episode. Very instructive.
I wonder when you are “doing math” if you should have a single example with many expectations. It seems like it would reduce the amount of signal to boilerplate. I am thinking something along the lines of this: Multiple expect invocations inside a single test example · GitHub
The problem that I see with this is that if there is a test failure it is not immediately obvious where the failure occurred. Perhaps it could be solved by a custom matcher or something.