arbitrary data can be done by pBFT replication algorithms, which don't have to be secured by tokens (or can be bonded to bitcoin). I personally think it's kinda sad that Nakamoto Consensus gives us 50.0001% defense against tampering and people think that going back to pBFT 2/3rd majority is a step forward.

Also, arbitrary data needs to be segmented by type or size or geographical or conceptual proximity to be cost effective. The whole reason why Bitcoin is a global database is because it's for a Global Currency. Data is not the same. Data has a definite set of interested users and thus such replication is pointlessly inefficient.