You get perfect parallelization. No need to separate tx into non-conflicting batches. With EIP-7928, every transaction can be executed independently of another.
While I agree that this is an interesting concept, it feels like we’re still kinda far from such a world. DAS, as done with blobs is a great role model for what we should strive for on the EL as well. zkVMs will further change how we think about related concepts, so, still rather difficult to tell how things evolve. Execution sharding has huge potential for scaling and is definitely on the list.