[jasomill]

You can make this arbitrarily difficult by moving more and more core functionality into the backend, to the point where the front end is little more than a bundle of cached assets for what's otherwise an online game. At which point, any replacement backend that reused these assets would inarguably be a derivative work infringing on the assets' copyright unless explicitly licensed.

You don't even have to go to this extreme to make conventional piracy all but intractable. As a concrete example, take Civilization 6, and suppose it had been released as a client/server application where the server handled all enemy AI. While creating functional but inequivalent replacement backends may not be terribly difficult — and might even lead to an interesting alternative AI ecosystem — reproducing the precise behavior of the vanilla AI via "black box" reverse engineering would require considerably more effort than cracking an offline game or reimplementing a backend that acts as a mere license server.

Compared to a traditional, fully offline model, moving large portions of a single-player game online would increase both upfront development costs and marginal cost and would be met with disapproval by a nontrivial fraction of potential customers who, for this reason, might choose not to purchase the game. It would also have a slightly smaller potential market to begin due to "always-on" Internet access being commonplace, but not universal.

Still, at this point, there are no real technical obstacles to developers adopting such a model.

In other words, I presume the forces keeping games "crackable" are primarily economic rather than technical.