The blockchain trilemma isn't real.
Certainly not remotely as real as something as rigorous and fundamental as the CAP theorem.
It's a loose framework on a dated blog post by Vitalik to justify sharding years ago.
It is disappointing how often it still gets misused by the crypto industry. Even Dankrad has publicly acknowledged that it's outdated and inaccurate.
Briefly, it states that you must have two out of three in: Scalability, Security, and Decentralization.
It is a useful construct for thinking about the extreme end states of each vector — i.e., if something is as fast as humanly possible, you have to sacrifice some amount of max decentralization to get there. Importantly, this doesn't mean that you have to be "centralized" to be fast — that's just not knowing basic reading comprehension, these aren't binary.
There are many current inefficiencies in blockchain architectures and system implementations. To take the blockchain trilemma literally is to assume that no further advancements are possible (i.e., you assume you can't tune software or existing hardware to get more performance out of it.
If you have a system and all you've done is improve the software to suck less while keeping everything the same — you've improved scalability while not reducing security or decentralization (in fact you probably have increased it because now people have more incentives to run the nodes).
The trilemma should be reformulated as: "You can't be as maximally performant as humanly possible while being as secure as humanly possible while also being as decentralized as humanly possible"
You can still be very decentralized, secure, and scalable. Simultaneously.
I understand nuance is hard in this industry, but thinking of these things as binary absolutes and then using this as some mental superiority complex for marketing is charlatanry.