We discussed this yesterday a little bit. It has several benefits but it makes resolving forks more complicated.
Do you mean
http://bitshares.org/documentation/group__dpos.html#dpos_conflict_resolution ?
I guess there won't be much differences between randomness one and 1,2,3,4 one.
Let's say the next miner is decided by RGN(N-2) % 100 and the results are like [2,3,1]. In this case, A,B,C is [2,3,1] and all the same rules in that doc work with [A:2,B:3,C:1] just like [A:1,B:2,C:3]. Notice that I used N-2 which means that it's already decided when current block is not generated yet and the next miner can be ready waiting to resolve the conflicts when it happens. In the worst case, we can use N-100 but I think N-3 could be fine as the possibility of both miners are down is low and in that case the next miner (decided by RGN of last 99) just picks up.
Edit:
In case that one miner doesn't produce block, it's just like ranking changes because we need to choose next next miner by using RNG of last 99. So there's really not much differences between the two from conflict resolving point of view.