@qhenkart Thanks for the question.
The short answer is: we believe there are better approaches to sparse-dense search than our current hybrid search implementation. You’ll see on our Considerations for serverless indexes page that the initial search is just based on the dense vectors. Cascading retrieval has been found to be much more performant for sparse-dense queries. Feel free to request Sparse Early Access (link on that blog post).
That said, we’re working on hybrid index improvements under-the-hood to improve recall. The documentation here is up-to-date, and we’re not scrapping this feature, but we do plan to release a better, more performant way to interact with sparse-dense vectors. More on this to come soon.