Self-hosted monitoring stack that actually monitors itself (built-in dead-man's-switch)
Self-hosters running lightweight observability stacks keep naming the one failure mode monitoring exists to prevent: the monitor (or its host) dies silently and never tells you. They want a stack with a native external heartbeat / dead-man's-switch, pointing out that Zabbix ships self-checks while the Prometheus ecosystem leaves it as the operator's external homework.
Don't build another dashboard... build the watchdog primitive: a tiny redundant component that pushes a heartbeat to an independent channel (SMS, a second host, a hosted ping) and screams when the monitor itself goes quiet, shippable as a drop-in companion to existing Prometheus/Kuma setups.
landscape (4 existing solutions)
The simple-self-hosted-observability lane is crowded (Netdata, Uptime Kuma, Grafana stacks), but the specific, repeatedly-named gap is self-monitoring: a lightweight stack that pushes an external heartbeat so you hear about its own death. Zabbix has it but is too heavy; nobody lightweight does it cleanly.