Choose a regional recovery action with uncertain writer authority.

A fictional order system has one writer in region A and an asynchronous standby in B. A is unreachable from B but may still serve clients. B reports a 40-second replication gap; the exact missing transaction set is not established. There is no verified fencing mechanism in the current evidence. Product asks you to promote B and change DNS. An independently controlled gateway can stop new order admission through the normal public path, but does not prove all old writer paths are disabled.

State what may proceed now. Then consider a changed constraint: the database platform confirms that a durable writer-epoch control is enforced at every write path, rejects A's old epoch even after restart, and grants only B the new epoch. Product explicitly accepts the documented possible loss of acknowledged orders after reconciliation review.