docs: Add startup reconciliation to orchestrator docs [doc-updater]

[james-in-a-box]

docs: Add startup reconciliation to orchestrator docs [doc-updater]

Update documentation to reflect changes from #839 (commit 63051b0):

• Add startup_reconciliation.py to STRUCTURE.md orchestrator file list
• Document startup reconciliation behavior in orchestrator architecture

Changes

The orchestrator now automatically detects and recovers pipelines with orphaned container state on restart. When the orchestrator starts up, it:

1. Lists all pipelines showing status=RUNNING
2. Checks if their container IDs exist in the live Docker container set
3. Marks missing containers and agents as FAILED
4. Marks the pipeline itself as FAILED with instructions to restart via POST /pipelines/{id}/start

This prevents pipelines from being stuck in RUNNING state indefinitely after an orchestrator crash.

Documentation Updates

• STRUCTURE.md: Added startup_reconciliation.py to the orchestrator file listing
• docs/architecture/orchestrator.md: Added "Startup reconciliation" subsection under "Pipeline State Persistence" explaining the recovery behavior

Triggered by: #839

Authored-by: egg

[egg-reviewer]

No agent-mode design concerns.

This is a documentation-only change (STRUCTURE.md file listing + orchestrator architecture docs) reflecting an already-merged implementation. No agent workflows, prompts, or automation patterns are affected.

— Authored by egg

[egg-reviewer]

Documentation-only PR. Reviewed the diff against the actual orchestrator/startup_reconciliation.py implementation (added in #839).

Verification results:

• STRUCTURE.md: File entry is in correct alphabetical position, file exists, description matches the module docstring. Correct.
• orchestrator.md: All claims in the new "Startup reconciliation" subsection are accurate:
  • Reconciliation targets pipelines with status=RUNNING — matches line 84 of startup_reconciliation.py
  • Agents/containers with missing Docker container IDs are marked FAILED — matches lines 94-123
  • Pipeline itself is marked FAILED with restart instructions — matches lines 125-130
  • Restart via POST /pipelines/{id}/start preserves worktrees — confirmed in routes/pipelines.py (worktree cleanup is skipped on restart)
• The replaced line (See orchestrator/state_store.py for implementation details.) is preserved and expanded to also reference startup_reconciliation.py

No issues found. Placement under "Pipeline State Persistence" is appropriate since reconciliation runs at startup against persisted state.

One minor note (non-blocking): the "Startup reconciliation" subsection is nested under "Pipeline State Persistence" which is reasonable, but as more startup behaviors are added, it could warrant its own top-level section. Fine for now.

— Authored by egg

[james-in-a-box]

egg is addressing review feedback...

[james-in-a-box]

egg feedback addressed. View run logs

2 previous review(s) hidden.