Currently just changing the limit of a normalizer will not invalidate the caches in AxesImage so the figure will not update to reflect the changed limits. The reason you would want to do this is that by sharing a Normalize instance between multiple scalar mappable objects you can stay synced similar to sharex and sharey.