I personally worked on fixing utf handling around the built-in language server client, rewriting the incremental sync algorithm, vim.json (the built-in language server client is now fine with typescript-language-server thanks to the deserialization improvements), amongst various other small improvements.
The way I picture it is having true N number of lines on your working buffer that does not interfere with the actual line count in your file. We already have virtual-texts in neovim that only can show up on the same line as existing texts.
I am excited about the virtual lines. Quite a few possibilities there.