This paper caught my attention:
"Our preliminary results show that Spider improves the number and volume of successful payments through the payment channel network. For a given amount of funds locked into the network, Spider is able to complete 10-75% more transactions amounting to a 10-45% increase in volume of transactions relative to SpeedyMurmurs  and SilentWhispers , two prior approaches to the path-discovery problem for payment channel networks. On an ISP-like topology, Spider also outperforms a classical max-flow based-approach by 5-15% on both the number and volume of successful transactions.
In summary, payment channel networks promise to be a key ingredient for scaling future blockchain systems. Their design presents exciting intellectual and engineering challenges, and we hope that this paper inspires more work in the networking community in this area."
I forward that hope.