Batch — the pattern everyone pretends they have outgrown

Events are fashionable; batches are effective. Payments arrive from billing in a nightly file of a few hundred thousand rows, and no event bus improves on that — the data is not urgent, the volume is high, and a batch is one API call where events would be a hundred thousand.

This lesson loads the file through Bulk API 2.0, handles partial failure row by row rather than batch by batch, and ends with the honest decision table: latency need, volume, and who owns the failure queue. Reading lesson two and this one back to back is the whole argument of the course — patterns are trade-offs, not fashions.