Three independent research tasks can run concurrently before a coordinator combines their results. Which design is appropriate?

1. Skip result validation because tasks ran concurrently
2. Parallel subworkflows with bounded concurrency and a defined aggregation step
3. Start unlimited workers without quota awareness
4. Add artificial dependencies between every task