RE: Steem Town Hall - Noon Eastern 16:00 UTC (3hours from now)
In your new scenario you would have 1/30 less voting power to overcome in order to take over.
Let X be the amount of SP required to take over a chain.
Let's use a really simplified example:
There are 80 Witnesses with the lowest one getting 1 SP and for each change there is a 1 SP rise.
Then if each user votes for one witness only there is 1+2+3+..+79+80 SP in total being used to vote.
That number is 80*81/2 (=3,240). In the opposite if every users uses all 30 votes there is (1/30) of that: 108 SP backing all those witnesses.
So our current state is 30 votes per user, so there must be at least 108 SP backing the witnesses. In order for someone to take over the chain, the user only needs to have 78 SP (or actually 77.001 SP) and he can vote 17 times and get sock puppets and take over.
So in this minimal case of 78 SP in this scenario ( and I haven't really been rigorous I admit in accounting for all distribution curves ) if we switch to 1 vote per SP, then the top witness doesn't get so much. Unless the distribution is changed, instead of 80 SP behind the top witness, then there could only be 2.667 SP. The 17th spot would be at 2.567 SP. Now, the user that has 78 SP can simply divide his vote by 20, 78/20 = 3.9 SP. Now that's enough to take the all 20 witness spots.
So in the 1 SP = 1 vote scenario: There is 108 SP backing the witnesses. If users change their habits? These 108 SP could change voting habits. Vote in 20 at 3 each. That's combining 60 SP together. To prevent hard-forks you just need 8 SP voting up contrarian witnesses.
Back in the current 30-votes per user, there is 108 SP. We can vote in 20 at 3 SP each but we can vote for those 20, 20 times each. With only 60 SP it's not enough. Indeed you would need 80 SP voting for 20 witnesses in order to defeat a Justin Sun. You need 60 SP to prevent hard forks.
At first it seems the proposal just weakens everyone by a factor of thirty, but by playing out scenarios we can see there are advantages to this proposed system for those who would have liked to see Steem consensus rules change less. Someone like myself would prefer this system.
in the real world there is some overlap between witness votes but actually (pre hostile takeover) there was not as much overlap between witness votes as your scenario assumes. For instance, not a single top witness had both the Freedom and Blocktrades votes, which were the two largest votes pre Stinc stake being used to vote. But as you point out, even in the very simplified and unrealistic scenario you present, being able to concentrate witness votes on just 4 witnesses to prevent a hardfork is actually a strong measure to deter hostile takeover.