I think that LfW is perfect as it is. I use it and recommend it every time someone asks me for a tip for beginners. I also test it whenever I can, and I'm willing to help even more, so I don't think it should have some radical change.
* Should Lua and LuaJIT v2.x be included?
LuaJIT impose some restrictions on modules so there will be less modules available for LfW to include if LuaJIT is the main interpreter. LuaJIT for me is a tool for advanced developers, LfW is a tool for everyone, that include beginners. And those advanced developers know how to rebuild any module to fit their needs. This already happen with IUP for instance, although the pre-compiled binaries are not compatible with LuaJIT, recompiling the IupLua binding with a simple define is enough to make it compatible (although leaving the internal Lua files exposed).
* What are the most important modules for the stripped down version of LfW?
Why you have to reduce the current number of modules? Since LfW is already stable, to update a module binary is not simple?
* Which packaging system should be used? LuaDist for building and LuaRocks for adding new libraries to users installs?
* Can this be made to be cross-platform? (I hope so, that is a personal goal)
Recently I used the Ubuntu and Fedora GUI based package installation systems. Both already have Lua and several modules available. I think that those sources will be preferred by users in a long term. Just a thought.
Anyway, if building from source code, it can be made portable. But again I think the charm of LfW is that it is great as it is for Windows.
Ryan, I don't get the need for another Holy Grail. You already have a Holy Grail... Despite the Run Time Library endless discussion, what are your real needs? For the end user, what will be the benefits from those changes?