You're not going to notice any lags if you use Emacs to work on toy projects. Try a non-trivial operation on a repository of respectable size, like linux.git: if you don't classify those operations as being "completely unusable", you're living in the dark ages.
Incorrect. Several proprietary IDEs use real parsers. Which is precisely why they're able to provide all sorts of magic functionality like "refactor". And no, regular expressions don't "work well" by a stretch: even the common modes like shell-script-mode, perl-mode (cperl-mode is slightly better), python-mode are riddled with bugs. If you look at the relatively new js2-mode, you'll notice that the parser is an Elisp port of the Rhino parser. Also, Semantic Bovinator from the CEDET project was recently merged into emacs, although it doesn't have major users yet. Precisely because regular expressions aren't "good enough".
The reason everything is implemented in Emacs Lisp is because there is no alternative: Elisp doesn't have an FFI to speak of. Otherwise, I don't see the problem with using existing parsers; unless your "point" is to take everyone back to the dark ages.
No matter how you look at it, there are more alternatives to Vim/Emacs than there were 30 years ago. A lot of people are happy with TextMate, Coda, Visual Studio, Eclipse. It's not like there's something special about the year 2013, but what I meant is: Emacs will become extinct, and that is inevitable. And yes, I'm hoping for some kind of programmable editor to take its place soon.