In his follow up post he talks about union combining identical results. If the author ever reads this, you can prevent that by using `union all` instead of just `union`. That won't remove duplicates.
Indeed, though his work-around of adding a path column is something is worth having anyway, because it makes it much easier to see where the generated output has come from.
It is forcing the engine to do more work though, because it demands a distinct result set, so it adds extra nodes to the plan to remove duplicates. If you know your query doesn't have duplicates, it's much smarter to not add an extra whole sort/distinct when totally unnecessary for correct results.
I'm fairly sure that a bakery doesn't produce the amounts of data where you need to worry about such things.
Eh, it's all relative. You can make some queries perform pretty terrible even with tiny data. It's something that has had measurable performance impacts on queries I've had to tune and deal with with over the years.
I don't know, when a baker needs a dozen of something, they tend to bake 13. Maybe things get out of hand.