I don’t think this explanation is very good. We could use a distribution where the expectation is defined and be back to square one. For example you can just tw

I don’t think this explanation is very good. We could use a distribution where the expectation is defined and be back to square one. For example you can just tweak his distribution so that the ratio of envelopes over the ratio of successive probabilities is < 1 (he has 10:2 which is >1).