HI @stamblerre! I can work on this issue especially as I am working now on #32339. The question is: is it really necessary to send HTTP request for each link in the file? I can see at least two possible issues here:

At first right now with implemented #32339 the documentLinks sometimes takes seconds to return the list of links. This is request for file with about 900 lines of code: [Trace - 4:46:35 PM] Received response 'textDocument/documentLink - (3)' in 3663ms.. Making the HTTP request for each link will increase this time dramatically.

The second issue is that I am not sure that developers expect network usage from language server especially during working with mobile internet. Making a bunch of HTTP calls on each documentLinks request can cause not only clogging of internet connection but also increased bills if you are working on mobile internet. I am not sure that scenario of usage mobile internet is wide and probably needs discussion.

Yeah, I am not sure that we should be doing this either. We could do some caching for links we've already seen, but I'm not sure that it's worth it. I filed this issue mostly to start a discussion, rather than to necessarily propose a solution.

[...] HTTP request [...]

FWIW, I think this is a bad idea. This is what happened when iTerm2 did something similar: https://nvd.nist.gov/vuln/detail/CVE-2015-9231.
EDIT: relevant discussion: https://news.ycombinator.com/item?id=15286956

Thanks for sharing. I think this issue is probably not very pressing, and I just wanted to open it to note that this has been mentioned, but I think we can close it for now. I'm certain this will come up on the future, and we can figure out a way to handle it then.