I haven’t really stress-tested it, but I would imagine that it scales well since (I) it can be configured to use Ripgrep to resolve searches and “goto links”, and (II) the “comefrom links” are cached on save. I haven’t used org-brain though, so it’s hard to compare :)