This solves a huge gap. Blue/green and canary rollout is critical for highly-available services.

We'd been half-solving it by doing a "sandbox" deploy, running a health check, then "promoting" to prod by then deploying to our prod environment. Automatically, without manual steps or verification, of course. But it's not true blue/green like this, doesn't account for accidental environment parity issues or drift, and doesn't accommodate canary rollout.

Puzzled by the relationship of SAM and Cloudformation in lambda ecosystem.

SAM seemed like an effort to create an open spec for this type of App but looks more and more bespoke to AWS.