I remember danabramov claiming more than once on twitter to people praising preact's bundle size, that it also came with a performance cost when it came to diffing and rendering because React had much more code to handle the many edge cases where preact would hit walls.
While it's a far-cry from a realistic test (few people are creating 1000 rows), React seems to be less performant across the board: https://krausest.github.io/js-framework-benchmark/2022/table...