Yes you eventually need to do the things you mention but probably less frequently than a normal application needs to add new columns or the like to support new use cases.

The article thesis is essentially make breaking changes as infrequently as possible. The easiest way to do that is never change your data but that’s a sure way to have your competitors crush you as you stagnate. The next best thing you can do is make sure existing producers and consumers are not impacted when you make changes. For most changes being made the advice in the article gives a set of things you can do to achieve this goal.

For times where your database itself is not scaling which are the types of things you’re mentioning, I think there are other things you can do to, if not eliminate backwards incompatibility, at least make the transition easier. For example fronting your DB via an API and gate all producers/consumers through that. If you’re frequently having to handle scaling issues perhaps it’s time to reevaluate your system design all together.