K. I just did some testing, and found some interesting results. (What, leaf did testing? Since when does he do that?) Anyway, here's what I found for IC Basics 20000-20005:
20k: current subaction
20k1: current action
20k2: ??? (see below)
20k3: seems to be previous action
20k4: pretty sure this one is always the value that correlates to the specific interrupt ID for whatever you're doing
20k5: ?!?!?! (see below)
For testing, what I did was place a graphic (external graphic 19 from the common file, if you're wondering... it was a very explosion-y experience) at the beginning of most relevant subactions that was of size equal to the IC-basic in question. Then I just played around with it in the game to see what I could observe (I used samus for the tests).
Now... for 20k2, most values seem fairly small (I think the largest is like 5 or something), but I can't really figure out the pattern. Most things you do on the ground (dashing, crouching, attacking, etc) appear to be size 2, but grabs, throws, and landing (regular landlag, aerial attack landlag, or special landlag) all seem to be size 1, while jumping seems to be size 3 and specials are size 4 (maybe... I didn't spend too long trying to figure out if the jump and special were different sizes, but I'm pretty sure they were). Something like that, anyway. Tough to tell on some of this... in the air, everything seems to be of the larger variety... I think there were two different sizes that I observed, meaning they'd be either 3 and 4, or 4 and 5. It's tough to say what this one is, really.
But 20k5... is just crazy. It's normally 1. Until you press something. Anything. Control stick. A button. Anything. Then... IT BECOMES HUGE. Like... the only way you could tell the graphic was supposed to be an explosion was the redish tint of the flash, which was actually more apparent in the background than the foreground. Needless to say, these values were sufficiently large that it was impossible to tell if there was any difference between them. I assume that there is some difference... since it doesn't really make sense to switch between 1 and a very large number... unless maybe if that very large number was FFFFFFFF, but even then, if it's just two values you'd think it would switch between 0 and 1. So there probably is some difference between the very large numbers, but until someone with a gecko gets in there, we won't be able to figure it out.
Well, actually, I can think of a way to test this... I'm just not going to. What you'd do is set RA-float[pickanumber] equal to IC-basic[20005], subtract a very large number from RA-float[pickanumber], then use RA-float[pickanumber] for the size of the graphic. But since dant is gonna be looking at IC-basics with gecko soon using this as a starting point, it'd prob just be best to leave it to him. Speaking of which, this is also the reason I won't be updating the OP with this info, since dant will probably be able to get much more complete information. Just thought I'd share with those interested =)