The presentation of blockchains as some kind of historical imperative would be downright Marx-like if it weren't for the primary difference that Marx put some thought into justifying his position. It's eminently possible to cryptographically secure software without lugging around an immutable distributed database because you're emotionally invested in the idea.
the blockchain is useful in solving double-spending problems in purely p2p applications. Aside from cryptocurrency, take for example name systems like namecoin or ENS: these systems need a way of reconciling who owns what, which involves synchronizing some data across the whole network.
It is inefficient, but the inefficiency seems to lie at some fundamental problem with p2p. Centralized systems need to do the same synchronization, but between fewer actors, and may outsource some of the verification for an exponential increase in speed.
blockchain isn't inefficient because it's p2p. it's inefficient because it assumes peers are untrustworthy and solves for that by imposing a proof of work, requiring cooperative peers to waste more electricity than bad actors.
assuming peers are trustworthy is assuming away the whole problem.
The problem isn't with p2p, but specifically with full-sync mutexes in p2p. There are petname systems that avoid the problem by avoiding the mutex itself, and there are centralized systems that avoid the p2p, but it comes down to zooko's trilemma.
