Add resources in each original region: Spinning up parallel monitoring nodes or services in each company's home region ensures true redundancy without moving user data outside its home zone. If one region goes down, regional replicas pick up the load, and data locality remains intact.
Use a network load balancer: A global or multi-region network load balancer can route user or agent traffic to the nearest healthy endpoint. This minimizes latency for a globally distributed user base and provides failover across regions, while respecting the requirement that each company's data stays on its own infrastructure.