The shape of the teeth is more reminiscent of the Ropean mk 1 than the Ropeman mk 2. I bought my mk 2 at Snell's in Chamonix because they said it would be better on icy ropes than the mk 1. I subsequently had the experience of jumaring up distinctly non-icy ropes in the UK using both a mk 1 and a mk 2. The mk 1 slipped a couple of times, and I was securely held by the mk 2 which has never slipped in any circumstances. I would be reluctant to swap my mk 2 for a mk 3 unless I was certain that it would grip the rope as reliably. The spikes on the mk 2 are arranged along a curved cam, which means that it would be far more difficult for the rope to escape.

if the teeth get completely caked it's not going to grip as well as the Mark 2.

I've always heard the mk2 was the best device to leave on a piece of pro when simulclimbing (whereas the mk1 or a tibloc would rip the sheath out in a big fall). How does the mk3 compare for this type of use?

You're definitely moving deep into the realms of improvised use with this, so it’s not going to figure in the manufacturer's instructions, and not an easy thing to test in lab conditions! However it is quite a well-known speed climbing technique, to reduce the risk under certain conditions, when moving ultra-fast on big walls. I would certainly be interested in conducting some tests, given enough rope and gadgets to destroy. However, it would be impossible to give definitive advice on this as there are so many variables, so “don’t try this at home, folks”. The idea is that the rope runs through a Ropeman clipped directly to a muli-directional runner (bolt/peg/thread) while the two climbers move simultaneously on a single rope. The idea is that the Ropeman protects the leader if the second falls off, because the second falls onto the Ropeman rather than the leader. However, to be honest all bets are off with this technique if either person falls! You also need optimal conditions, otherwise the rope drag can be a real problem for the leader, and end up slower than belaying each other properly. The Tibloc is a non-starter for this sort of tactic, in practice it needs a little finger squeeze to move from paying out mode to gripping, so if the second climber fell it would definitely strip the rope and probably fail to grip at all. The "word on the street" is that the Ropeman 1 would cut the rope at point of failure whereas the Ropeman 2 would strip the rope at point of failure. In either case you might get away with it if relatively little slack creeps into the system. Very hard to simul-climb at exactly the same speed though; in my experience it works best for aid climbing. So this technique basically reduces the risk from probably pulling off the leader, down to giving a fighting chance of survival – neither of which is the sort of thing would get a good quotation from a life insurance broker. On the whole I think the Ropeman 3 is probably the best (least bad?!) for this kind of technique, but in any case it’s based on a presumption of not falling off (so marginally safer than soloing, but survival still requires a pretty high luck factor if the second falls off). This is based on having smaller teeth than the Mark 1 so less likely to sever the rope, and producing less friction than either of its predecessors. Personally I prefer to shave time off the ascent by quicker change-overs and lightweight gear. It's a handy trick to have up your sleeve for the odd occasion, but regular use is not conducive to a long climbing career, regardless of which gadget you use.