The "write one why line per cloud" step is underrated. Most two-cloud setups drift because nobody can say which workloads belong where, so new stuff lands wherever the person deploying it is most comfortable. On Rule 1, one detail that trips people up: AWS cost allocation tags only show up in billing from the point they're activated, not retroactively, so it pays to activate them on day one even before the tag scheme is perfect. Matching lowercase keys across both clouds is a small thing that saves a lot of joins later.