TL;DR Based on an audit, I estimate about 100 of the 120+ P6 advent articles of the last 5 years remain 100% current; about 5 to 10 are fundamentally broken; most of the remainder have only trivial problems.
Update David Warring has begun working his way through the advent posts creating tests for the code in them
How many years of advent articles no longer work because they only ran on Pugs or Necza or Parrot and those are now broken?
(Fwiw I think some contributors to these projects would say they're not broken.)
Using the P6 advent search box to search thru the 5 years worth of advent articles:
3 articles match "Pugs". None of the articles in which Pugs is mentioned have Pugs specific code, so no breakage.
21 articles match "Niecza". There's a "contributing to P6" article that focuses on Niecza and C# code. Let's call that fundamentally broken. Two articles (tetris and the 2011 mandelbrot) use CLR libraries. The P6 code itself looks valid, so these articles might work fine if/when there's a Rakudo/.NET, but I'll call these fundamentally broken. The rest don't contain Niecza specific code or instructions.
15 articles match "Parrot". The two articles Reini Urban wrote last year are Parrot specific (threads, and low level encoding details). The rest don't contain Parrot specific code or instructions.
How many no longer run because the design of Perl-6 changed?
I'm not currently willing to go thru all 100+ articles. I've gone thru the first 10 articles for 2009 and 2011 and entered the examples in a current Rakudo. Afaict, all of the 2009 articles remain 100% valid and two of the 2011 articles generate a message (both trivial to fix):
So it's not worth your time to review a significant portion of the important documentation you keep claiming that people will be working on in 2013? That sounds about right, ralph. Hurry up and shovel something out the door everyone. Don't worry if it works in six months. That's why you don't have users.