A possible use case when I would like to analyze a huge C++ project with CodeChecker  and use clangd to help me with the jump to definition feature.
So, I am assume that the yml file could be relative big for extremely big projects and in this case I wouldn’t want to generate it for all files. It could be generated separately for translation units. And if I am requesting the locations for different definitions from different translation units the index yml file can be dynamically reloaded.
This is just an idea but I hope I formulate it clear and easy to understand.