Tuning CloudFront TTL from Lambda@Edge is a sharp lever: rewrite Cache-Control on the response path and the POP suddenly keeps HTML longer — or shorter — than the origin intended. The operational question is whether every behavior and every region still agree on the key after a deploy, and whether a TTL change alone is enough or you still need a targeted invalidation. Self-hosted edge fleets face the same shape once inventory spans nodes: one place to see what each node stored and to coordinate purge. We have been building CDNCA (AGPL-3.0) around CDN nodes and a plugin marketplace from one control plane; marketplace features roll out release by release. After you moved TTL into Lambda@Edge, did header rewrites or an explicit invalidation clear stale viewers faster on the next publish?