The incident has been resolved.

Two cluster nodes are failing, leading to process restarts and degraded availability. We're looking into it.