The "hybrid" as used in this article referred to a partition scheme that combines a "Lomuto partitioning like algorithm" together with a "Hoare partitioning like algorithm".
This amplifies the the strength of each scheme and mitigates the weakness. Ie. Lomuto does too many swaps, but naturally fits a branchfree implementation. Hoare is frugal with swaps, but is rather branchy. The "hybrid" here inherits the frugality of swaps from Hoare's scheme, while benefits from the branchfree implementation of Lomuto scheme.
What you seem to describe here is the classical and well-known hybridization of divide and conquer algorithms. This is also used and implicitly implied when branchless bubble sort is discussed. But the "hybrid" as used in the article wasn't meant to refer to this.
This amplifies the the strength of each scheme and mitigates the weakness. Ie. Lomuto does too many swaps, but naturally fits a branchfree implementation. Hoare is frugal with swaps, but is rather branchy. The "hybrid" here inherits the frugality of swaps from Hoare's scheme, while benefits from the branchfree implementation of Lomuto scheme.
What you seem to describe here is the classical and well-known hybridization of divide and conquer algorithms. This is also used and implicitly implied when branchless bubble sort is discussed. But the "hybrid" as used in the article wasn't meant to refer to this.