feat: add graph workflow fallback and retry metadata

docs/workflows/WORKFLOW_DELIVERY_CHECKLIST.md

@@ -28,9 +28,9 @@
 
 - [x] Workflow context introduced
 - [x] Node outputs are persisted and reusable
-- [ ] Graph runtime supports legacy fallback
+- [x] Graph runtime supports legacy fallback
 - [ ] `legacy`, `graph`, and `shadow` modes exist
-- Progress: Graph dispatch now executes the extracted bridge nodes (`order_line_setup`, `resolve_template`, `material_map_resolve`, `auto_populate_materials`, `glb_bbox`) synchronously, persists per-node outputs/logs/durations onto `WorkflowNodeResult`, and continues to queue render/export nodes through Celery without changing the legacy preset dispatcher.
+- Progress: Graph dispatch now persists retry/failure policy metadata per node, retries bridge-node failures within the configured attempt budget, and the production order-line dispatcher can opt into explicit graph mode with a hard fallback back to legacy dispatch if graph execution preparation or runtime fails.
 
 ### Phase 5
 

docs/workflows/WORKFLOW_IMPLEMENTATION_BACKLOG.md

@@ -73,9 +73,9 @@
 - `E4-T1` Introduce `WorkflowContext`. `completed`
 - `E4-T2` Refactor executor to process nodes against context and node outputs. `completed`
 - `E4-T3` Persist node-level run records, logs, timings, and outputs. `completed`
-- `E4-T4` Support retry and failure policies.
+- `E4-T4` Support retry and failure policies. `completed`
 - `E4-T5` Add execution mode switch: `legacy`, `graph`, `shadow`.
-- `E4-T6` Add hard fallback to legacy dispatch on graph failure.
+- `E4-T6` Add hard fallback to legacy dispatch on graph failure. `completed`
 
 ## Epic 5: Editor Parity