The trilemma and the choice not to brute-force it
Ethereum could have chased scale the blunt way, raising the gas limit, shrinking block time, or requiring beefier validator hardware so the base layer alone processes more transactions per second. Every version of that path runs into the same wall known as the blockchain trilemma: a chain can optimize hard for at most two of decentralization, security, and scale before the third gives way. Push throughput up by demanding more from node operators, and fewer people can afford to run a validator, which concentrates control and quietly erodes the decentralization that makes the chain worth trusting in the first place.
Ethereum's answer was to refuse that trade. Instead of scaling L1 execution, it scales by keeping L1 deliberately modest, cheap enough to run on consumer hardware, and pushing the actual transaction execution off-chain onto rollups. L1 stops trying to be the place where everything happens and becomes the place everything else answers to: a highly secure, decentralized settlement and data availability layer that rollups post their results back to. This is the rollup-centric roadmap, and it's less a feature than a philosophy about where compromise is acceptable and where it isn't.