Lumen's fulfillment team uses a basket extract rather than the full order table. Some records arrived before their item list was populated. Your report must keep those parent orders where the task asks for every basket, and exclude them only when the requested output is item grain.
This set mixes array length, safe positions, correlated aggregates, EXISTS, UNNEST, joins and windows. None of the exercises requires a new construct. The main challenge is choosing where the expansion belongs. A shipping charge belongs to the order; multiplying it by the number of item rows would overstate the fulfillment cost.
Use the fields and aliases requested in each task. Where a report asks for product units, sum quantities. Where it asks for orders, count distinct order IDs. The final task adds a chronological running total to daily basket counts without changing the date grain.