Reviewed the staged Factory campaign prototype against the current implementation without editing the mono worktree.
Verified all six histo scopes, the viseg scope, case IDs, completion predicates, effective worker parameters, batch sizing, resources, worker source behavior, and controller-to-worker payloads.
Confirmed each campaign evaluation creates a fresh vxData client.
Ran the focused Factory tests and staged-diff whitespace check successfully; no review findings remained.
Lessons Learned: Pitfalls
The controller sends partial parameter dictionaries and lets worker Pydantic schemas fill defaults, so comparing raw payload dictionaries would falsely suggest parameter drift. Effective validated parameters and actual runner usage are the meaningful comparison.
The new one-worker-per-campaign admission cap deliberately changes backlog scheduling, while batch sizes and worker execution behavior remain unchanged.
Lessons Learned: Improvements
Keep a focused compatibility test that validates every campaign payload with its worker schema and asserts stable producer IDs and case UID fixtures.
A differential test over representative histo and viseg data frames would make future completion-predicate migrations easier to verify.
