Test group 2 has landed. I think most of you will be pleased with the results. After a 30 z-level fall, 90% were immediately dead, with 80% fully dismembered (with parts flying as high as 2 levels above the floor of the test chamber).
It was very odd, of the 9 dead, 4 dwarves had announcements of their death (`Subject 2B' Darosiden. Test Subject has died after colliding with an obstacle) and no combat report, 1 had no announcement but did have a combat report ("The test Subject slams into an obstacle and blows apart" with no other hits reported), 3 had no announcement nor combat report, and 1 had both an announcement and a combat report (9 hits, including skull -> brain). The survivor had no announcement (yet) and did have a combat report (8 hits, including upper spinal damage that should be fatal very quickly).
Ah, I just noticed that the dead test subjects that had no announcements are reported in the 'u' list as missing, while those that had announcement s are reported as deceased. That makes sense. I expect they'll get announcements as soon as the cleaning crew comes to tidy up the test chamber. I'm not sure why some are still reported as missing, given that there is a surviving test subject on the floor of the test chamber (he's even conscious, for the moment). I don't see a pattern involving how close other falling subjects were. One of the missing had 2 test subjects still in flight on the bottom level when he finished the test, so they should have seen him.
Two dwarves failed to disassemble, and took 8 or 9 combat hits, shattering bones and doing fatal damage (either immediate death, or mortal wounds).
I would guess that if you dropped enough dwarves, you might get a few that were lucky enough to not disassemble and even survive the multiple hits. But more testing would be needed in order to even make an estimate of what the survival rate would be. It seems extremely unlikely that any would be able to walk/crawl away, or even remain conscious. A peasant with a sock standing by near the bottom would be sufficient to finish them off.
Three dwarves did collide with the walls on the way down. This slowed them considerably and resulted in them being the last three to finish the test procedure. The two test subjects that failed to disassemble on impact were both involved in collisions on the way down, and I would guess that they didn't disassemble because they weren't moving as fast as the others. However, one that hit a wall on the way down did disassemble upon completing the test, so using the walls for braking isn't a guarantee of an intact corpse (nice try, though).
The fastest subject finished the test in 48 ticks, the slowest that avoided interaction with the walls finished in 56 ticks. The three that slowed down to put red graffiti on the walls took 61, 66 and 66 ticks. Their next of kin would be sent cleaning bills for the graffiti, except that we were careful to make sure that no one in the test group had any relatives that weren't also in the test group.
We plan to repeat the test with goblin test subjects, as soon as some show up and volunteer. The cages welcome mat is already in place for them. I may also try doing some testing using hatches instead of a retracting bridge to see if that will give more consistency in lateral as well as vertical accelerations. I may also did the test chamber deeper, it's in a spot that can be extended all the way down to 119 z-levels deep. We could do 120 if we want a magma landing, but I don't think we'd learn a whole lot from that. It would make the cleaning crew's job easier, though.
TL;DR: 30 z-levels does seem to be sufficient for 90+% kills, with 80% disassembled on impact. \o/