Reliability: external heartbeat / dead-man's-switch for total-down detection

[brocla]

Backlog: Reliability & durability

Problem

FileMill runs unattended on a personal laptop. If it crashes, the machine reboots for updates, or the tunnel drops, nobody knows. Self-sent email alerts (the error-alerting plan) structurally cannot report "the process is dead / never started / the box is offline" — the alert would have to come from the thing that's down.

Proposed solution

Add an external heartbeat / dead-man's-switch: FileMill pings a third-party monitor (e.g. healthchecks.io) on a timer while healthy. If the ping stops, the monitor emails/alerts you. This is the only mechanism that catches total-down, and it's independent of FileMill and Mailgun.

Scope

• A small ticker in filemill run that GETs the monitor URL every few minutes (config: heartbeat_url).
• Mostly external setup (create the check, set the grace period).
• Complements auto-restart (gets it back up) and catch-up Durability: catch up on missed mail at startup (pull from Mailgun storage) #2 (recovers missed mail): heartbeat tells you when even those failed.

Notes

• Keep it dependency-light and non-fatal: a failed heartbeat ping is logged, never crashes the worker.