A single complex query spills heavily and runs slowly on a small warehouse. Which compute change should be tested first?

1. Increase the warehouse size
2. Reduce the warehouse size without measurement
3. Add only more clusters to handle the same single query
4. Create another identical idle warehouse