System health

Each check measures an effect the system is supposed to produce — rows landing in a table, a timestamp advancing, a queue draining — rather than whether a job reported success. Failing checks also post to Slack.