A Warehouse Bill Doubled Overnight — Diagnose the MPP Cost Spike

Your team runs a Snowflake warehouse. Overnight compute cost for the "analytics" virtual warehouse roughly doubled, with no change to dashboard usage patterns or user count. Investigating, you find:

• A new nightly job was added that runs SELECT * FROM events WHERE event_type = 'click' against a 4-billion-row events table that is partitioned (clustered) by event_date, not event_type.
• The events table has grown 3x in the last quarter but was never re-clustered or reviewed.

1. Explain, using the MPP/micro-partition pruning concept, why this specific query is expensive relative to what it "looks like" it should cost.
2. Propose two concrete fixes and explain the mechanism by which each one reduces cost.
3. A teammate proposes simply setting the virtual warehouse to a smaller size to cut cost. Would that fix the actual problem? Explain what it would and wouldn't change.