language-servers using clangd/ccls/... are definitely useful, but quite resource heavy. for example, each of these tools seem to starting new threads per file (!) and there are no knobs to not do that. i don't really understand this rationale at all. yes, i have seen this exact behavior with both clangd and ccls. oftentimes, the memory in these processes balloon to some godawful numbers (more with clangd than ccls), necessitating a kill.

moreover, this might be an unpopular opinion, but mixing any regex based tool (ripgrep/... come to mind) with language-server f.e. because the language server does not really find what you are looking for, or does not do that fast enough, are major points against it. if you already have language-server running, regex based tools should not be required at all.

i don't really understand the reason for sql'ization of code searches at all. it is not a 'natural' interface. typical usage is to see 'who calls this function', 'where is the definition at' of this function etc. etc.