We've seen cases where temporary table cleanup fails, but the errors often surface well after the original issue, making them hard to trace. To improve debugging, we should add temp table usage to DRT. This could help surface failures while we still have relevant context like logs.

This issue tracks adding a new DRT operation that creates and exercises temporary tables.

The operation could:

• Create a temporary table or view
• Add DML for the temporary table
• Sleep for a short period to keep the session active (to simulate cleanup paths triggered by session end vs. node crash)
• Disconnect to trigger cleanup logic

Jira issue: CRDB-53334

Epic CRDB-17128