Right, I think this somewhat corresponds to the "how long it takes a policy to reach a stable loop" (the "distance to loop" metric), which we used in our experiments.

What did you use your coherence definition for?