How would you handle two users updating the same record in an EF Core API?

In production I choose optimistic concurrency when conflicts are uncommon and the read-to-write interval includes user think-time, where pessimistic locks would be harmful. The token detects a conflicting row update; it does not protect a multi-row invariant such as total capacity, which still needs a transaction, constraint, conditional write, or stronger isolation. Deletion and missing-row cases are mapped deliberately so they are not confused with a normal edit conflict. Tests use two contexts against a real relational database, assert the generated token predicate, and cover reload, merge, explicit overwrite, and authorization after conflict. Metrics distinguish expected conflicts from failures and reveal workflows where contention is high enough to require a different aggregate or command design.