A large Redshift fact table is queried primarily by narrow event_date ranges. The date distribution and workload are stable, and scans read many unnecessary blocks. Which physical design should be evaluated?

1. A suitable sort key led by event_date, or automatic sort optimization, with effective table maintenance.
2. Use only a distribution key on event_date and assume it provides the same block ordering as a sort key.
3. A sort key on a random value unrelated to any filter or join.
4. A larger query result LIMIT and no changes to physical design or predicates.