I am worried about a scenario where we are running 100% on AWS and relying on AWS NTP and something goes wonky with their NTP service, and the cluster starts evicting nodes to where we lose quorum. Are there any gates that can be set for the number of nodes which can be evicted? Furthermore, I would rather the cluster go read-only than become inaccessible / unresponsive when losing quorum.
@fat0 @ik_zelf, I created an issue for tracking: https://github.com/cockroachdb/cockroach/issues/26688. Please add more details there. Anyone coming across this issue via search down the road, please feel free to do the same!