So you did! That's pretty hilarious that you didn't even realize you discovered something this big, the level really does drag on huh... That's probably why they decided to end it early, in the end. Turns out the missing lives were under our noses the whole time!

While hunting down some strange additional extra lives that weren't accounted for previously, I noticed that all these extra lives seemed to extend past the end of the level. What I didn't expect is that there was actually more level behind the end of level trigger! It's almost completely functional (there's a non-functioning electricity gate, and the moving platform control boards are ridiculously slippery) and yet just... not accessible normally. Perhaps the developers figured the level was too long and hard already, and slapped down an end of level trigger a couple of rooms early as a last second stopgap measure, which would explain the level ending before even reaching the final pink platform.

Video of this extra piece of level here.

To visit it yourself, load up 8-A (don't move Yellow), open Cheat Engine, search for the array of bytes A2 70 92 C1 14 A1 C7 C4 77 A8 F0 C2 and replace it with 02 C9 82 45 5F A1 16 47 19 5C 90 43 Keep in mind that going any further backwards than this will hit the end of level trigger because it is very big.

2-C's case is strange, since you seem to be hitting and breaking a door element itself - I assume the data for this door was flagged incorrectly somehow, since all the other doors seem to be fine.

5-A is especially weird since certain segments of it run at very high speed on my computer, much faster than normal, but not even the whole level - unfortunately I only have the one computer to test on so I can't do much testing in these matters.

But in much more exciting news, I managed to track down the missing lives in 8-A, and I think it warrants its own thread!

I ran some code analysis on the executable and couldn't find any code paths that reference these strings. That's not to completely rule out access to the editor, but since I was able to find code paths for the cheat codes and whatnot, it doesn't look good...

Most of these lives aren't actually real. These unfound entries appear to simply be the name for the actor that the level creator used in the level editor (some BoxLife elements are named BoxLife1, BoxLife#5, etc; the actor name comes 54 bytes before the actual actor object in the .LEV format). I guess whoever worked on the earlier levels was better at giving custom names to the actors and got lazy with the later ones.

8-A appears to be the exception, this level has names for all actors and DOES actually have 18 lives! I'm not sure where all of them could be since the coordinate system is a little strange, but by modifying their positions in the .lev file to the starting location I was able to get all of them to spawn without any tricks like 1-C. 8-C I'm less certain about, but those may be real as well since I'm not seeing the same duplicated "BoxLife" strings in the name field.

Premods aren't really an issue IMO, due to the 1/4 second delay between entering song select and being able to pick a song you can actually select mods without losing any time if you're fast on the keyboard. Whether this should be required as some display of skill is up for debate though.

Seeing results screen is a tricky one, with properly timed escape presses you can skip past it before it becomes visible, while still passing through it so it submits and generates a local score. I prefer this definition myself since otherwise there's no difference between quitting after the last combo-giving note and quitting after the last hitobject ends, but it's certainly not as clear cut to me.

How is finishing a map defined? likean00b's current #1 score doesn't actually "finish" the map in that it doesn't submit to the server or generate a score on local ranking, so as far as osu! knows it's actually just a mid-map quit.