We don't have direct S3 access. Entra-AWS identity friction makes that a non-starter for now, which means passively reading external tables from S3 isn't viable; not unless we're committing to full truncate-and-reload cycles on billion-row tables. We're not. So we built around

I'm generally a best-practices person. I think there's wisdom in the consensus positions that the data governance community has built up over decades, and I try not to be contrarian for its own sake. But some recommendations read like they were written for an organization that