An image workflow contains several managed service steps, conditional branches, and recoverable failures. Operations needs per-execution history and explicit retries without a custom workflow engine. Which design fits?

1. Use one S3 lifecycle rule to execute conditional business branches.
2. Use Step Functions with service integrations, bounded retries, and explicit failure handling.
3. Use a DNS weighted record as the workflow state store.
4. Chain fire-and-forget Lambda invocations and infer state only from log text.