Great point, Ahmet. In our current design, the fingerprint's artifact_sha actually tracks the downstream code module where the check executed (proration-engine's sha:9f82d1a), not the entrypoint invoice-orchestrator. So updating or rolling back that downstream artifact forces a cache miss and triggers a fresh check. That said, your point about binding the entire traversal chain—including intermediate hop versions and route edges—makes total sense. Hashing the full path topology would prevent stale hits if an intermediate service routes differently without changing the leaf artifact. Definitely planning to roll that into the next iteration. Thanks for reading and sharing that!
