Conductors should only initiate take-over when a conductor joins or leaves the cluster. In other words, when any conductor service starts or stops, it may trigger a re-balancing, and while other conductors should "heal" there could be a race leading to this behaviour as a result of the cluster changing.
Was this behaviour being observed even when no ironic-conductor services had been started or stopped for some time?
Conductors should only initiate take-over when a conductor joins or leaves the cluster. In other words, when any conductor service starts or stops, it may trigger a re-balancing, and while other conductors should "heal" there could be a race leading to this behaviour as a result of the cluster changing.
Was this behaviour being observed even when no ironic-conductor services had been started or stopped for some time?