They are somewhat right. I have built these feature flag/kill switch kind of things and they rarely get tested. Over time it might not even work or have other side effects.

On the other hand, a product like Twitter having some content moderation filter seems very likely.