I fully believe there are differences in the emulator loading times. I have been testing out RNG manipulation lately and part of it includes doing something similar to a TAS. What I've noticed is that the program I use doesn't always get everything exactly right. I believe this is due to the emulator loading the game fractions of a second faster or slower. Although I say all of this, I don't believe these small differences are an issue at this point. If the speedruns for this game get to a level where fractions of seconds are making the difference in the end, then the issue should be further addressed.

I have been working on understanding how the game decides what items go in which boxes. Obviously this depends on what place the player is at the time of breaking the box though. One thing I've discovered so far was that all items in a race are determined at the beginning of each race. So on lap one of a race, the item that will be in the last box on the last lap is already set and will not change. I haven't been able to figure out whether these are sets of items that are used multiple times or if each box is randomly generated at the beginning. I have found other small bits of interesting information such as a theory that it's also possible to manipulate the AI to not use items. I have a recording that shows this but still need to do more testing. I just wanted to let everyone know what I've found so far so they can possibly build off of it.

Thanks for the quick response, I completely agree. I'll check the timing and other elements to ensure that everything is exactly the same. You can look forward to seeing my runs in the near future.

Are reproduction carts allowed as long as loading times are the same?