Each bar is one automated DB health probe (~1/min). Green = healthy, red = failed.
Probed in-process per worker. Ops are paged automatically on a sustained outage. Live data from /v1/status.