What the resume-in-pizza-boxes channel offers over, say, the resumes-in-email channel, is a lot fewer resumes. You've got some intention signalling in the RIPB version (it's expensive in terms of time and other committments), but low informational signalling otherwise as to overt qualifications, outside some general concepts (hustle, motivation, time availability / opportunity costs, charisma, confidence, or simply having heard of the trick somewhere).
There are a bunch of selection hueristics whose primary benefit is they eliminate a large component of the choice process. You'd do fairly well in simply randomly sampling from amongst the submissions, or applying some very rudimentary filters to eliminate the most obviously bad choices.
This is why mechanisms such as racial, gender, age, and cultural bias are so common. They're easy to apply, aren't entirely wrong (relatively homogeneous groups tend to operate better than nonhomogeneous ones, though there are many arguments about resilience and creativity/diversity), and hugely reduce the decisionmaking costs of the selector.
Unfortunately, knowledge / infotech / creative work hinges hugely on intangibles which are quite difficult to effectively communicate, especially to someone (say, an HR flunky) not intimately familiar with the field.
Toss in a ton of employer biases and control incentives (e.g., anti-poaching, anti-union, and similar concerns) and the situation starts getting messy quickly. Sure, there might be some alterative opportunities created as a result, but these tend to have far less commercial/financial potential and viability.