# TASK 4.1: Implement Normalize-Load Orchestration And Upstream Raw-Batch Resolution

## Summary

Implement the tenant-day `normalize_load` entry point and the ETL/Warehouse orchestration flow that starts one `raw -> load` run for Brink `Sales2.GetOrders`.

This subtask turns the replay boundary from `TASK2.md` and the lifecycle contract from `TASK3.md` into a concrete execution path that resolves the upstream raw batch, builds normalization context, and prepares load artifact persistence.

## Dependencies

- [TASK 2](./TASK2.md) - extract raw XML
- [TASK 3](./TASK3.md) - batch lifecycle, checkpoints, and failure packaging

## Files to Create

### 1. Normalize-load orchestration services

Create or finalize the services responsible for one tenant-day `normalize_load` execution.

Recommended logical services:

- `BrinkPosNormalizeLoadJobs`
- `BrinkPosRawBatchResolver`
- `BrinkPosNormalizeLoadOrchestrator`
- `BrinkPosNormalizationContextFactory`

## Required Behavior

- child normalize-load accepts a `BrinkPosNormalizeLoadJobInput`-style contract
- explicit `ExtractBatchId` wins when provided
- otherwise the resolver loads the current successful `extract_raw` batch for the same tenant-day key
- the upstream raw batch must be successful and replayable
- normalization context is built from persisted raw artifact metadata rather than in-memory DTOs
- the orchestration layer passes only normalization context and raw artifact references into the Brink normalization boundary

## Testing

### Unit tests

Add unit tests for:

- explicit upstream raw-batch selection
- implicit upstream raw-batch selection through current-state lookup
- rejection of failed or invalid upstream raw batches
- normalization-context creation from raw artifact metadata
- orchestration handoff from child job to normalization execution

## Acceptance Criteria

- [x] Child normalize-load job entry exists
- [x] Upstream raw-batch resolution exists for explicit and implicit selection paths
- [x] Failed or invalid upstream raw batches are rejected
- [x] Normalization context is built from persisted raw metadata
- [x] Unit tests cover upstream resolution and orchestration behavior

## Notes

- Keep this subtask focused on stage orchestration readiness.
- Do not implement relational decomposition or Parquet persistence here; those belong to later subtasks in `TASK 4`.