Two large Redshift tables are frequently joined on customer_id. The join key has a well-distributed population, and the workload analysis shows network redistribution is a major cost. Which distribution approach should the engineer evaluate?

1. Use DISTSTYLE ALL for both very large tables without evaluating the replication cost.
2. Collocate the tables using a suitable KEY distribution on customer_id, or evaluate automatic distribution against the workload.
3. Choose unrelated distribution keys for the two tables and expect identical column names in the query to collocate the data.
4. Use a constant distribution-key value for every row to ensure all joins use one slice.