Redundancy as a deliberate design choice
Defense in depth layers multiple, independent controls — network boundaries, authentication, authorization checks, encryption, monitoring — so that bypassing or defeating any single one still leaves other layers standing between an attacker and the actual resource. It's a deliberate acknowledgment that every control has some failure mode, and that the response to that fact is redundancy, not a search for one control good enough to not need it.
For authorization specifically, this often means pairing a strong access-control decision with monitoring that would catch misuse if the decision was somehow wrong or bypassed — not because the decision is expected to fail, but because a second, independent layer catching a failure the first layer didn't is exactly the point.
The trap of over-trusting the strongest layer
A system with one exceptionally strong control and comparatively weak layers behind it is, paradoxically, more fragile than one with several good-but-imperfect layers — because the strong control becomes a single point that, once bypassed, faces nothing else meaningful in its way. Investing all confidence in the best control undermines the actual purpose of having several.