You are building a Cosmos DB solution deployed to multiple Azure regions. Requirements: read operations must never observe writes out of order, and you want to maximize read concurrency in all regions. Which Cosmos DB consistency level satisfies these requirements?

Consistent Prefix ensures that clients never see writes out of order. If a user reads version A of a document, they will never subsequently read an older version of that document (like version A-1). It allows for high read concurrency and low latency, as it doesn't enforce strict global ordering or require all replicas to be up-to-date before serving reads. Session consistency would be too strong as it guarantees monotonic reads, monotonic writes, read-your-own-writes, and write-follows-reads within a single client session, which is more than "never observe writes out of order" and can impact global read concurrency. Eventual consistency is too weak as it provides no ordering guarantees. Bounded staleness allows for some out-of-order reads within a defined window.