200MB blocks every 10 mins would compete with paypal's throughput, and wouldn't significantly reduce decentralization. A couple cryptocurrencies of this size, perhaps merge-mined, with atomic swaps between them could cover most of the world's need for transactions without the security and complexity of payment-channel based solutions. This seems to line up with satoshi's original intentions.
Mobile users could just shard, prune, or use SPV on a typical residential internet connection. Meanwhile merchants and exchanges can afford ~$200/yr in hard drives to store the whole chain (It is only about 10TB). When you consider that an efficient entry-level bitcoin miner is in the many thousands of dollars, the cost of a couple hard drives a year is not very significant in the grand scheme of the centralization of the network.
You sacrifice a lot with payment channels. You have to be online to accept payments, you need to pre-fund a channel (that you pay expensive fees on in the case of a small blockchain), you need to route through other nodes with enough liquidity, you need to occasionally go online to prevent the channel from being insecurely closed.
Even if it was the case that the economy should primarily rely on the lightning network, it would still be complemented by a large adjustable blocksize for reasons mentioned in this article and the lightning network whitepaper. The current status quo is unmistakably bad for the security of the network.