You've done a great job, and I am in no way intending to be critical. My only concern was the claim of accuracy. It is an outstanding simulation of the original, and I commend you for your work on it.
However, leaving out the bugs, for example, makes it not accurate. As for the timer, I don't see how you could guarantee the frequency in a javascript VM. It could be similar (by continually testing how inaccurate it is and adjusting the frequency of subsequent iterations), granted, but this is not accuracy (it is emulation). As for memory, etc, many of these older games had "death screens" where a buffer overflows and simple crashes the game, preventing you from continuing past a certain point - that too would be very difficult to accurately model in javascript. The list could go on and on. And this is in no way critical of you, or what you have done - it is simply statement of fact about accuracy!
As an aside, Billy Mitchell (the first person to ever get a perfect pacman score) was one of the stars in the movie The King of Kong.
Again, I think what you have done is AWESOME. But it is not accurate (and likely could not be).
I probably phrased this incorrectly, hence the downvotes. These arcade machines have specific timers, odd hardware characteristics, hard memory limits, etc, that are likely impossible to replicate in software accurately - in addition to the software algorythms they use.
In king of kong, you will hear how a blob of solder in the right location will change the characteristics of the game to invalidate the entry for world record status.
Simply looking at the framerate counter at the bottom you can see that this isn't in any way "accurate". By looking at the github repo, you can even see the author has written about some of the inaccuracies - timing, collision detection, randomization, and bugs.
If you have problems disengaging from your work, I strongly recommend having a go at archery. Focusing on nothing but a spot of gold way down the field does wonders for clearing your head :)