I am working on the right formulas for mute/follow effects on holding or releasing Stake power via modulating effective reputation score, to avoid flamewars, some of it is going to be hard to model because of its complexity, but it's complexity is due to concurrency rather than complex sequencing.

It should not be sequence dependent so as individual accounts change the parameters only one operation is required per change. Virtual operations I guess they are called, storing a transient state derived from deterministic data coming from the database state.

I don't see how there can be any patterning discovered in comments that can positively determine humans. The best turing test is other humans, and the best turing test of all, is many other humans. That's why this model puts the tools to submit these judgements in the hands of the users rather than trying to guess it with heuristics.