Hey Steelcrocodile! Sorry I missed the post, not used to people posting on here lol.

Your questions are good ones, and these kinds of examples are why "glitchless" is a very difficult thing to nail down.

I've done my rough take on a glitchless run a bit ago linked below. The main idea is to ban obvious "large" skips (such as in 4-1 using the horses to kill the gohmas instead of the bow or the 2-2 skip you mentioned) along with the obvious glitches in the game. As for what you mentioned in #3, this exploit is mostly fine (can use it in 7-2 to get to the fairy easier) I just didn't use it when it skipped a lot like in your example of 2-2.

The only glitch that I decided to allow is recoil FLG (floating link glitch). This is pretty much impossible to avoid (almost any clank you get with a sword hit will activate this) so I decided to allow it. It doesn't really affect too much in the run (the biggest is probably the 6-3 boss). I can't see any objectively correct way to go about it, I just revolved the rules around not skipping large portions and logical to some extent.

1. Feel free to run whatever you find the most interesting and what you have the most fun with. As for the leaderboards, usually categories get added out of interest from multiple people. There are a lot of categories one could make for this game, but it would just clutter up everything. But again, just because it isn't on here, don't be shy about running whatever you want.

2. Those are the only 3 emulators that are allowed. (Bizhawk is really the same as Gambatte)

3. You are not allowed to bind one key to multiple inputs. One key has to correspond to only 1 button.

4. In terms of glitches and tricks, 1.0 (US and JP) has a trick called super jump and a few other things (I think) while the other versions do not (or a nerfed version of them). The text in this game is suuuuper slow, so JP saves quite a bit of time from text. Also in the JP version there are some NPCs where you can completely skip their text by pressing B.

After being not lazy and downloading an emu with frame advance, I looked at the various ICS setups to see if they were "random" like mine above.

First, testing my method, buffering up+select out of the SQ menu 3 pixels away from the instrument would oscillate between failing and succeeding every consecutive frame. e.g. (frame 1 starts on the SQ menu)

Trial 1

frame 1: up+select+A result:failure

Trial 2

frame 1:up+select frame 2: up+select+A result:success

Trial 3

frame 1:up+select frame 2:up+select frame 3: up+select+A result:failure

Trial 4

frame 1:up+select frame 2:up+select frame 3:up+select frame 4: up+select+A result:success

Also, other methods similar to this (getting 1 pixel from the instrument and pressing map and direction a few frames after) have a similar issue unsurprisingly. I specifically inputted select and left two frames later 1 pixel to the right of the instrument. This gave the same results as above.

However, trying the "yolo jump" method, I think this is 100% consistent assuming you pause on the correct frame. Now, I'm not sure if this frame is always the same but there seems to always exist a frame to pause when you land on the instrument. I only did a few trials of this to see if it had similar issues as the other methods so I could be wrong.

Sorry for the long video, but I found a "setup" (probably old) for ICS that doesn't work and I don't understand why. Basically, from what I understand, after unpausing or after getting out of the S+Q menu, buffering direction + map moves you a set amount of pixels. For some reason this is really inconsistent with ICS and I don't really understand. I tried to troubleshoot why this was happening and concluded it came down to when you unpaused/came out of the SQ menu. I'm just wondering what others make of this.