It's not on a whim; I've actually thought for a while now that "non-partisan IP license store" is the perfect use-case for a blockchain—one much more fitting a blockchain's particular system of incentives than the exchange of monetary tokens.
An account for holding IP licenses is, at minimum, just a bag of arbitrary tokens (UUIDs, say) signed by IP owners. To make them useful, the tokens might be, say, SHAs of files they are considered licenses to.
The files themselves don't have to be attached to the system anywhere; in fact, the files don't even have to be available from the IP owners. Holding such a token effectively says "I, the IP owner, promise not to sue you for having a copy of this file on your hard drive."
All the files themselves could be available on BitTorrent or whatever else—the tokens might-as-well-be scrips to participate in decriminalized piracy, much as doctor's notes allow one to participate in decriminalized marijuana consumption. But you can imagine an ecosystem of clients (or maybe OS-level features) that can collate licenses to IP-holder CDNs to get those files, maybe transmitting a zero-knowledge proof of the possession of the license to the CDN server in exchange for the file; and will remove shell-level access to the file when the license is removed from the store. (The file could be retained in an OS-protected cache and such in case the user regains a license to it; it just has to be made inaccessible to the user.)
Now, consider the current products and services that a complete implementation of this sort of system would displace: any App Store (Apple's) or Game Store (e.g. Steam), any Music or Video store, potentially most streaming services, etc. A proper implementation would make "having content available to access" just an OS-built-in transparent side-effect of having the proper licenses. Tons and tons of middle-men would be put out.
So, imagine building this thing in any way that's even slightly centralized. Imagine, just as a slight step away from a blockchain, that this was a Ripple/Stellar type thing: a set of ledgers extended such that each ledger account can hold a bag of bitstrings, rather than a bag of {currency_code, amount} pairs.
Wouldn't it be in this service's interest to gain adoption by partnering with these IP middle-men? They'd say "let's get Valve to make it so that people's Steam games are listed in here, and then people can just move the licenses around using our wallet app! Win-win!"
The potential partners, afraid of being displaced and commoditized, would go along, but at the same time, try to take control of the system. They would want the technology provider to allow them to impose business restrictions on how and when users could transfer things, or even charge a cost payable to the partner for doing so.
The result wouldn't be a distributed zero-trust ledger of IP licenses at all; it'd just be a meta-store interface, with each store able to enforce its own policies on its licenses.
Now, you could posit that there's some perfect organization (maybe the FSF? GNU Taler is a thing) that wouldn't fall for these tricks, even as its network absolutely fails to gain traction because nobody wants to be displaced by it.
But it's much easier to just solve the problem by not having an organization running the network at all, so that there's "nobody there" to decide to partner with the IP owners or the current middle-men. Take the decision away; turn off the incentive structure. Just make it a blockchain.
(The other convenient thing about such a structure being a blockchain, though, is the scaling story. A company like Stellar builds out a network and then has to pay to run its network until people start to join and peer with it and overwhelm the seed nodes; if Stellar were to run out of runway at any point, it would almost certainly be at the start, the costs of maintaining the seed network draining them dry before adoption picks up. A completely distributed ledger is initially run by idealist hobbyists, who will continue to run it without charging it rent each month until such a time as the system gains popularity enough to comprise commercially-run nodes as well. It's like parents taking care of their kids for free; it lets the network grow up until it's ready to support itself.)
---
...now, all that being said, yes, you can eliminate the "blockchain" part of this if you flip the problem on its head. If instead of the representation of ownership being the computed ledger account holder of a given token, the tokens themselves are self-describing documents that are signed and re-signed by each party as the token changes hands—then you can store these tokens anywhere: put them in a DHT, stick them in your OS's keychain store, wherever. People can just carry one to another computer via Sneakernet if they like.
This quickly gets into problems with verifying that a license hasn't since been moved or revoked, though, which starts to look like an "ask the IP holder if it still wants you to be holding its license" sort of problem—like CA revocation is today. In fact, the whole system in that case would look like the current infrastructure for code-signing—and that's a problem, because that sort of system privileges IP owners (who can run CRL servers) at the expense of consumers (who can't or won't), making the whole thing just turn back into trusting the IP owner to say who currently owns a thing, and getting the IP owner to perform license transfers on your behalf, and all the rent-extracting problems that those two things generate.
With a blockchain, everybody is up-to-date on who currently owns that thing. You don't have to think about key revocations or expiries; you don't have to worry about stale grants; each machine just computes current ownership after every new block, and if something is now gone from your account, all your devices know it (or know that they're offline and can't trust their database of ownership assertions.)
But, if you have a system for broadcasting those self-describing documents—something that where everybody will be able to synchronize with it and then receive timely updates from it (Usenet, maybe?)—then yes, you get the same properties as a blockchain. But you've also just effectively recreated a blockchain. (And while you're recreating a blockchain, you might want that "proof-of-work" thing in there, too, so that people can't flood your newsgroup-ledger with pointless messages.)