This fixes #61. If there are a number of Bagger instances on the same host in the middle of the list, we will detect and rotate accordingly.
The new logic is to rotate by the maximum number of bagger instances hosted on the same physical host. In the case where someone has more than half running on the same host, then some failure tolerance will be lost, but that's their fault. I suppose if it becomes critical we can detect and warn in that case, but I doubt it will be a serious problem at scale.
This fixes #61. If there are a number of Bagger instances on the same host in the middle of the list, we will detect and rotate accordingly.
The new logic is to rotate by the maximum number of bagger instances hosted on the same physical host. In the case where someone has more than half running on the same host, then some failure tolerance will be lost, but that's their fault. I suppose if it becomes critical we can detect and warn in that case, but I doubt it will be a serious problem at scale.