One question I've not seen answered is whether we are expecting to stick to 'only' 101 delegates. Surely that limited number is a liability for any attack.. and inversely might not be enough to include everyone that might be due support. For example, if BitShares goes large, are you really limited to only a handful of people in every quarter of the globe, or to those hiring others directly? With more delegates, 100% would be a smaller fraction too. Given that running a delegate is trivial, perhaps it's worth considering delegates and payments separately.
This is a good question that sort of got lost in this thread. Let me take a shot at it:
Bytemaster did an analysis of what the right number of delegates should be. His argument was that going from one delegate to two doubles the security, going from 100 to 101 increases it by less than one percent.
But the increase in cost of each delegate is linear. So it is a simple matter of diminishing returns - from a security standpoint.
Also, since there are always many other candidates waiting to fill a slot, attacking any one delegate has little effect. Someone else will pop up to take her place. So the actual number of people defending against attacks is much higher than 101. We have engineered a global whack-a-mole system where delegates will tend to pop up in whatever jurisdictions are friendly to them.
Another consideration was the size of burden placed on the shareholder to determine who gets their votes. The primary voting task is to determine who should be trusted. Once that is done, we trust those people with "trivial" tasks like selecting the correct delegate software and running it reliably. In the future, we will tend to elect trusted managers who can oversee multi-person projects, one task of which is running the correct delegate software reliability.
So, doubling the number of people to assess for trustworthiness would demand an impossible task from everyone. We would all defer to some small number of consumer advocate delegates to tell us who to vote for. And thus, not really generate any more decentralization for the delta cost.
Seems like 101 is challenging enough. So it was chosen as the "sweet spot".