Is your feature request related to a problem? Please describe.
I know there were latency concerns in the past with memberlist and splitted brain problems with large clusters.
But maybe we can revisit this and find a way to enable it, maybe if only experimental/supported in small clusters of 100 nodes.
Describe the solution you'd like
I should be able to allow HA tracker with memberlist without needing etcd or consul
Additional context
Users have to run an etcd cluster / consul to run cortex ha-tracker