Doing it through TCP reminds me of an idea I had once about doing a mod consisting in using multiple UT servers as a single persistent cluster, and which I would use both UDP and TCP, with a central controller in some other language.
I didn't really dig much yet on the link classes from UT, but if you meant to use sockets, in Linux at least it would be best to use a local Unix socket instead for that communication, as it doesn't suffer from the needless complexity and lag (comparatively speaking) which TCP introduces.
But I don't really know if UT supports Unix sockets at UScript level.
Adding to this, real support for a secondary programming language in UT is possible I believe, such as Lua.
But the entire API would need to be written natively in C++ to create an interface between Unreal Engine and Lua itself in the same way you have UScript with the engine.
Would it be worth the trouble? Maybe, maybe not.
In terms of speed, if LuaJIT is used for instance, we're talking about one of the fastest programming languages all around in existence, it gets extremely close to native performance, making it far faster than UScript, plus you get to be able to go beyond that and be able implement and fix things you wouldn't be able otherwise.
However, replicating the entirety of the classes to Lua, and since Lua is weak-typed and follows a completely different paradigm, it might prove to be too much of an hassle to be feasible.
For instance, Just Cause 3 has a multiplayer mod made exclusively by the community, and this mod can be developed for in NodeJS (server-side) and by using CEF (client-side), thus being both Javascript, for both the frontend and backend.
And thus far their API has almost nothing, not even the bare essential things to even create a mutator or a gametype, but you can pretty much design a beautiful UI through HTML and CSS due to the usage of CEF (Chromium Embed Framework).
But they didn't have access to public headers like we did, they had to reverse engineer a lot of it, and maybe had a few hints from the actual developers of the game as they ended up endorsing the mod, so I believe we can do more powerful things.
However, at the end of the day all of this would require native modules to interface with each respective language, and unless there was an effective and trusted way of distributing these, all of this development would be pretty much for naught. Might be an interesting project to take on as a hobby, but as something more to develop more serious mods for UT, I just don't see that happening, unfortunately.
Personally I am planning something however, but on the very long run (something to start maybe several months or a year from now, if at all), and it's not the usage of an existing outside language, but rather kinda creating my own in a way with the current UScript itself. Something more limited than UScript, but more accessible and easier to do things on by players, without necessarily any performance hits or minimal ones through actual compilation and tokenization through normal objects.
But there's still a lot I need to do first in and outside of UT first for this to be even worth the trouble, so it's not something I am going to start anytime soon.
But if someone else comes along and does something similar, or brings another language in the meanwhile, I am certainly going to look forward to it.