Whoa, join a game a week after everyone decides to get rid of the one thing that actually caught your interest in the game… O blessed art thou, circumstance. Well, small loss if they go through with it and I disappear, I guess. 
I get that its support is probably a pain, and indeed so many people are doing things in hard-code. But the problem isn’t that everyone’s doing stuff in the native code and therefore Lua support is irrelevant… the problem is that everyone’s doing stuff in the native code when they shouldn’t. Esoteric, purpose-driven scripting or behaviour should be externalised; the stuff that’s hard-coded in a game should almost always be limited solely to core mechanics and platform support. Things like door pulleys, etc. should be externalised as a matter of best practice. The vast majority of the mapgen stuff is being externalised already, and the only holdouts are, no surprises, scripted functionality.
Lua would also be pretty much essential for NPC behaviour and dialogue, unless one can honestly assert they’re willing to put in the work on a JSON-based solution that could rival, say, Fallout 1’s level of complexity. (Which Black Isle used scripts for, I might add. =)) At the very least, it should be easy to foresee that, once people stop worrying over the gameplay mechanics and start concentrating on the late game, numerous amounts of scripted behaviour will be critical. Lua is the glue that affixes the game to the sandbox. It’s a great sandbox game, but its longevity is limited to an in-game year at best, and the only meaningful way of introducing dungeons, conversations, quests, and unique artifacts will be with a script engine.
“Not being used” as a justification for removal is a self-fulfilling prophecy: it will indeed never get used if you get rid of it. Rather than throwing out the baby with the bathwater, concentrate on showcasing it and discouraging improper use of the engine for stuff that really should be maintained in soft-code. Pretty much every time someone says “I don’t think this should go into the mainline” is a pro for Lua support. =) If you give people actual time to use the functionality it adds, and provide documentation that isn’t as obtuse as raw compiler-generated output, we’ll be more likely to see interesting mods cropping up here (like those of the Japanese community).
[edit]Just saw the thread linked here on Github debunking the “time” part, but wanted to mention the game currently attracts only programmers and sandbox-players because there isn’t an actual plot going on – shift the focus onto a campaign (presumably, downloadable independent campaigns) and you’re far more likely to see writers and designers showing up as well, who will be clamoring for anything that doesn’t require them to get down and dirty with the core code, and yet will be deeply and profoundly disappointed with the limitations of JSON or the (relatively) astronomical turnaround time on requesting engine features needed for their questlines. Thus, my “time” argument is not based on “give it time”, but rather “wait for the time where it becomes relevant”. No one will use the Lua functionality right now because the game just isn’t far enough along for it to matter, but that’s no incentive to get rid of it… that’s incentive to keep it in now so it doesn’t become an absolute bitch to bring back in later. If this game will only ever remain a zombie sandbox, then this of course no longer matters.
That all said, there are more than a few examples of iuse features in those Japanese mods that would make a not-insubstantial number of people pissed if support were removed entirely. =)
