You guys, this discussion is way more fascinating than just "should we fix old bugs"?
When I watched the video, he noted that the fix was to switch an "F0" byte to a "D0" byte. Those translate to BNE (D0, "Branch Not Equal") and BEQ (F0, "Branch Equal"). The 2 instructions are used in the exact same way, but one is positive and one is negative. The syntax of 6502 (or any machine language for that matter) is not always the same as spoken language though. It's not always clear whether you should use BEQ or BNE. Sometimes, when you describe what your code does in "Plain English", you have to translate it by flipping all of the positives to negative, or vice versa.
Now, that's fascinating enough, but let's add another layer onto it. When you translate English to Japanese, you get the same effect. Sometimes things that would be expressed as a negative in English need to be expressed as a positive in Japanese, and vice versa. Example, if you asked a child, "You didn't hit your sister, did you?" an English child would say "No", responding to the "Did you" portion of the question to say that they didn't. However, a Japanese child would say "Yes" to indicate that your initial statement "You didn't hit your sister" is correct. I learned this the hard way as an ESL teacher. Note that most machine language is initially based on English syntax. This was a problem back in the day, as it was harder for Japanese speakers to learn the correct syntax for their code.
Putting that all together, you have a perfect recipe for mixing up BEQ/BNE. I can look at this video and say, "Yeah, it's pretty clear both what the bug is, and how it happened." So do we preserve the mistake out of a sense of nostalgia, or do we acknowledge that it was clearly a mistake, patch it, and move on? My vote would be to patch it. No toggle switches; no options menu. Just change it to what it was clearly supposed to be from the beginning.