And yes, it's a chicken/egg scenario -- how do you prove that IPFS adoption would negate the need for CDNs, without widespread IPFS adoption? But currently this isn't some amazing Slashdot proof solution. They needed to setup 3 servers and a whole bunch of software when a single server + free CDN would've accomplished the same thing, faster and cheaper. It's presented as though CDN+IPFS is why this works so great, when really it's just CDN and IPFS just makes it more complicated.
If the goal is to learn, try out new tech, break dependence on CDNs, then this is great, but a single note in the article that this is a learning opportunity and it can be done easier without IPFS would've saved the author a lot of back and forth in the comments here.