Compute recovery starts with the business operation and the state it needs. An image can help recreate a machine, while durable records, credentials and dependencies need their own recovery design. A second running copy doesn't automatically preserve the valid state before an accidental application change.
These original scenarios distinguish Availability Domains, fault domains and the resource-compartment boundary. They also connect recovery time and tolerated data loss with tests a real operator can perform. The browser doesn't provision a standby, restore a disk or measure recovery duration. Use the decisions to define what evidence would demonstrate recovery of the actual service. A resource record with a familiar name is only one part of that evidence.
Separate recreating compute from restoring usable service and data.