Pick the right scope
Choose a narrow set of workloads to hedge: critical APIs or auth, not your entire stack.
Document what failure you are mitigating and how fast you need to move traffic.
Multi-cloud only pays off if it shortens customer downtime, not just adds optionality.
Make workloads portable
Standardize images, IaC, and secrets so deployments are reproducible across providers.
Keep shared observability with Watch.Dog and a single SLO view for both clouds.
Prove failover works
Run quarterly traffic-shift drills with synthetic monitoring to measure RTO and error rates.
Align DNS TTLs, certificates, and data replication so cutovers are boring.