The fact that an alternative perspective that proves simple, effective, and popular in alternative communities is now popularized from stars in the field of deep learning rather than the alternative communities is exactly this.
In fact, the resources available to the stars is much different than what these other communities may have access to such that they may never be able to demonstrate value themselves. It's not very often you have a 720 core machine to run these tests. Resource constrained science exists all over from chemistry, physics, and biotech to computer science. The rich get richer as they say.
However it's also a great instance of solid science to test alternative approaches and I don't mean to take anything away from anyone, but simply to point to a case where that it is extremely difficult for people to challenge stars in the field.
Evolution Strategies as a Scalable Alternative to Reinforcement Learning
by Tim Salimans, Jonathan Ho, Xi Chen, Ilya Sutskever
This seems like counter-evidence to the idea that people are unable to challenge the stars of the dominant paradigm.
As a specific example look at evolutionary optimization. It's existed for decades, has a bit of a non mainstream cult following, but now the leaders of the field are finding they can beat the fanciest deep reinforcement learning approaches with it. https://blog.openai.com/evolution-strategies/