When is Eventual Consistency Okay?

I’d say CA would generally require a single node setup, which might not be practical in real applications. With something like CRDTs, you might argue for a multi-node system that remains available while still retaining eventual consistency, but it gets complex.

Eventual consistency can actually be desirable in several cases. For example, analytics pipelines don’t need real-time data—who cares about the exact page hits right this second? Non-analytics apps like live polls or trending indicators can also tolerate slight inaccuracies, making them more user-friendly. Even systems like DNS registration or inventory tracking often work on eventual consistency. It's surprising, but these systems can function just fine without immediate accuracy.

Eventual consistency works well when a user is interacting with data that's only relevant to their session and can’t be updated by others. For instance, in offline modes, where the system operates like a local machine, partitioning becomes less critical.

A classic example is when you upload a video to YouTube. It's not available everywhere instantly, but the experience is smoother for users. Similarly, Reddit counts upvotes that might take a bit to reflect accurately but prioritize page speed. In contrast, immediate accuracy is crucial for things like buying tickets on Ticketmaster. If someone is trying to snag the last seat, that needs to be updated right away.

I’m not sure why you're asking about eventual consistency when talking about CA; CA setups are pretty rare. You really can't have a foolproof system against network failures, and losing data isn’t something anyone wants. Eventual consistency is often seen in high-traffic events like social media comments, state stores for availability (like DynamoDB), shopping carts, email, and more.