I am in the position to have more and more use-cases for AutoLoader. This might be an obvious question but if I understand correctly, Christian's use-case without the landing stage could still be an auto-loader scenario if the extracts from source system always land into a dedicated storage area (let that be ADLS or S3 or ...). That way you still don't have your own version of the files (with its pros and cons) but you can utilise AutoLoader's directory tracking ability.

My current requirement is to ingest data from an external S3 bucket and similarly try to evaluate the various approaches.