How would you upgrade the Presto engine version in a production watsonx.data environment with minimal disruption?

The safest approach is blue-green engine replacement: provision the new engine version; run validation queries; then update routing to direct traffic to the new engine; finally decommission the old engine after confirming stability. The incorrect options ("Modify the Presto binary directly on the coordinator node during a maintenance window", "Delete the existing Presto engine; which automatically triggers an upgrade on restart", "Upgrade is not possible; the entire watsonx.data instance must be reprovisioned") are distractors that don't fully capture the concept described.