Thanks.
I think the random seed does not really solve the problem here because there is no ground truth for the approximation; you would rather have analytic expressions as reference. Therefore, the isclose() approach sounds interesting. Though I guess its not clear how to define what’s close enough.