I've posed this question casually to viewers in my channel a few times before -- I've often wondered if the point along the girder section from which you take off has any impact upon how wide of a jump is possible? For example, if by moving forward one pixel Jumpman also moves UP one pixel (while running uphill onto the next girder section), it would make sense to me that the entire trajectory of Jumpman would be higher if taking off from this higher point. This means that at the tail end of the jump, just before landing, Jumpman would be slightly higher and/or he might land further out. Also, if you were to take off on that first higher pixel, it would seem to me that you are allowing the barrels the longest amount of time to roll LOWER (as they roll downhill onto lower girder sections) before Jumpman decends upon them on the other side of the grouping.
I think it would be a fun experiment if someone (not me) would try to set up a MAME savestate of a very wide, but possibly jumpable barrel grouping and try jumping off of each pixel of a girder section and chart out the results to see if it might possibly matter on the fringes of super wide groupings. You'd want to use the memory address features within MAME to "see" the x/y location of the barrels in question so that we could know the exact width of the grouping and (also looking at Jumpman's x/y location) (see the DK code for memory locations) you could be sure that you were jumping from the latest possible moment (I'd imagine playing the game frame-by-frame would be useful). It would be interesting to know if any other factors come into play as well, such as "winding up" a jump by running the other direction and then quickly reversing and immediately jumping the other direction, or if it matters which Jumpman sprite is drawn on screen at the moment the Jump button is pressed, or if Jumpman's "lost" frame while running plays a roll in this as well (see the excellent Spring analysis which I'm too lazy to link to which shows that if Jumpman's approach begins on a "lost" frame, it can mean the difference between success and death on certain spring combos).
If it turned out that the location along a girder section DOES matter, I believe this is something that could be practiced and eventually executed when necessary.
