[tabtab]

> allows potentially complex ETL statements to be handled within a single merge statement. I'm not saying that's always the best practice, but if things are structured in the right way, it may be.

I wonder how often one has the ideal schemas for such. The advantage of an ETL is that it's probably easier to debug, log, and trouble-shoot problems. One-Giant-Do-All-Command is often a recipe for headaches. Being able to "X-Ray" the intermediate pipes really helps troubleshooting.

The X-ray issue is one reason I have trouble debugging functional programming. Maybe there's a way to X-ray better, but I haven't found it yet.