Batching is how generative production stops being a person at a keyboard and becomes a pipeline. One set of conditioning, one credit ceiling, one run log, one review at the end rather than a decision after every render.
The ordering matters more than people expect. Run the hardest shots first, not the ones at the top of the shot list: hard shots reveal the real acceptance rate within the first hour, while there is still time to change the approach. Easy shots make you optimistic and tell you nothing.
A batch that produces no ledger entry is not a batch, it is a session.
How big should a batch be?
Small enough that a failure pattern is caught before it has been repeated fifty times. In practice that means batching by shot type rather than by deliverable, and reviewing between batches.
Why run the hardest shots first?
Because they tell you your real acceptance rate while the budget is still uncommitted. Discovering a fifteen per cent pass rate on the last batch of the week is discovering it too late.