This is my fault. I never thought about someone painting a ring for the endpoints map. In that case, it breaks the algorithm. The algorithm that makes the fiber field kind of tries to find paths from the black endpoints to the white endpoints. Inside the ring, all paths lead to the same color, so there's no path between the endpoints. It ends up setting the direction fiber to [0,0,0], which later triggers an error about non-normalized fiber directions (you should see this in the script editor when you do this, telling you which vertex is bad).
For now, my recommendation is to either add a point of the other color somewhere inside of that ring, or don't paint rings
For the future, I'm not really sure what should happen in this case. What result are you expecting for the fiber direction inside of that ring? I could try to make that happen. Or I can detect this case and print a more descriptive error message ("error: you painted an island of grey in your endpoint map" or whatever).