Sure thing! I will be dropping by tour posts if they are about blockchain core theory. It is true, all partitioned systems are bounded to give up either availability or consistency... the CAP theorem establishes a theoretical limit, but when designing consensus algorithms you can perceive that there are layers built within the system which individual availability/consistency trade-offs. For example, steemit only has 21 witnesses capable of validating blocks. By reducing the size of the system network latency, distributed stage updates, and other features are more easily handled.

But what really decides whether a consensus mechanism is successful or not, is its ability to economically reward trustworthy behaviors and to punish byzantine ones.

cya.