Apologies @Rohan that I didn't explain why Sloppy Quorum wouldn't work for this case.

Sloppy Quorum works well for handling requests when the entire quorum isn't available/is partitioned. However, it also increases the chances that if the client connects to a partitioned node, it will get stale value, which breaks our linearizability guarantee. So Sloppy Quorum favours availability, at the sacrifice of strong consistency.