How is search implemented with fast ranking factors?

Question

Algorithms

From the Yandex article habrahabr.ru/company/yandex/blog/204282 I took the terminology "fast and slow ranking factors".

Factors are of two types: fast and slow. Slow factors are calculated once and are assigned exclusively to the document. Quick are calculated for the document along with the search query.

With slow factors, everything is very clear, we are building an inverse index. All this is already implemented in Sphinx, ElasticSearch, etc. Where can I learn more about the implementation of search with fast factors? Or maybe there are ready-made open source implementations of the algorithm?
For example, there are 10M documents and 100K search queries. It's not a problem to keep a separate reverse index for documents and separately decreasing/increasing coefficients for the query-document link. But how to perform a search at the same time, combining these data, keeping within 0.1 seconds, I can’t imagine.