I view Memo as a nice proof-of-concept, but long form / high volume storage will be better served by other blockchains or solutions long term, and then this tech will fade out. My 2c.
I actually tried the "multiple blockchains" strategy with data in mind. It introduces more attack vectors than just IPFS. It might work in some cases tho.
Maybe we need a way to incentive archival nodes. If eventually there's a lot of data, we could develop something like swarm nodes: parts of the blockchain stored in multiple nodes.