> Instead of keeping deleted data in the same tables from which it was deleted from, there can be a new relation specifically for storing all deleted data The disadvantage of this is that if you ever do want to access this "deleted" data, e.g. in admin or compliance tools, you now have to do it in two different ways, one way for the main data and a different way in case the data has been "deleted". The article assert…
IMO it's worth distinguishing between (A) some kind of "click to undelete" feature versus (B) simply having that old-data conveniently exposed for a developer to manually-edit things or craft database-change scripts.
In practice I've only ever seen the latter get used, because it requires a developer to figure out how the heck to get "the parts that matter" back while preserving the integrity of other newer data and obeying certain business-rules.