A query has a substantial slowdown after a recent change. Which evidence best identifies expensive execution stages before tuning?

1. The query profile and execution metrics for the affected query
2. Only the warehouse idle timeout
3. Only the source table row counts
4. Only the elapsed time shown for one run