There’s been some interesting back and forth on Twitter about this. It’s still quite early but Roasbeef seems to take the point of view that optimal routing while neat is not necessary. 

With a sufficiently large network you can begin using probabilistic routing quite successfully I imagine, e.g. A* algorithm. 

But if a market emerges around finding efficient routes trampolines on the network have incentive to find efficiencies wherever possible.

k00b

People might not realize the importance of this work yet.
If the lightning network keeps growing at its current pace, this improved routing could save us from many tears down the road.

Another benefit of this work: it will make bigger payments over lightning very feasible.