Accelerating Bitcoin's Transaction Processing

Finally something about bitcoin that's actual hard CS and not fuzzy speculation or wishful thinking.

From a very superficial reading I understand that this paper proposes to solve forks in the blockchain by choosing the subtree with the most children (instead of the largest height, like in the current protocol). The authors claim that this avoids wasting (much of?) the computational power invested into blocks that are eventually rejected.

The meat of the paper starts at page 18.