Just my 2 cents as someone who has used emulators for a couple of decades -- I think Billy must have had help from someone who knows a thing or two about
MAME. In the dispute thread it was pointed out that any CS student could have commented out the lines in the source code and I agree with that, but that is only part of the story.
Tool-assisted superplays were in their infancy in 2004, I'm not even aware of any
arcade runs from that time, the few videos that existed back then were all for the NES using an emulator called Famtasia. You'd have to be totally brazen to even attempt a TAS with the early versions of MAME, considering 1) MAME used samples for the audio up until
version 0.116 in 2007 so that would be a giveaway; 2) the timing was slightly off, which would also be obvious to competitive players; and 3) MAME did not support re-recording until 2010.
MAME did have save states early on but they were not perfect, if you tried to string together a series of states into one input file there was always the chance it would desync, you could edit the .INP in a hex editor but again that's not something I'd expect Billy to be able to do on his own. If I were trying to make a high score run using the tools from back then I'd probably opt for splicing the video segments together digitally rather than trying to build the perfect .INP.
At any rate I'm wondering if Billy or one of his pals ever reached out to anyone on the MAME development
team (or maybe the MAME Action Replay Page?) to ask how to make MAME look more like a legitimate arcade performance. For example the site forums.marpirc.net has posts on recording MAME input files dating back to 2002, from the name I'd presume there was also an IRC channel back then. I'd be very curious if any of the
people involved in recording DK ever got an e-mail or message back then because I just can't see Billy or any of the people in his circle figuring out on their own that they needed to set the refresh rate to 60.6 or underclock the CPU.