The fleet, live
The pulse, in public.
Every box in the fleet reports two things to one place: a heartbeat when it boots, and the result of its integration-test suite when it runs. Elsewhere on this site those show up as screenshots of an admin. Here they are live, read straight off logger-service, refreshing on their own. This is the same page a client gets pointed at their own fleet — the receipt that the system shows its own state instead of staying silent until it breaks.
Reading the fleet…
What you are looking at.
Two independent receipts, both read server-side from a single box. The test receipts are the newest integration-test run recorded for each service — suites that exercise a real database and real service wires, not mocks, so a green count is evidence the promised workflow held, not that the test agreed with itself. The fleet heartbeats are each box reporting that it booted: presence is the signal, and the timestamp is the last deploy or restart, so an old stamp is a long, stable uptime rather than a problem.
Nothing here pages anyone — the pulse is read, not pushed, on purpose. The fuller account of which surfaces exist, who reads them, and where the system is still blind is on built to show its pulse; the map of the boxes themselves is the architecture.
And the tests are themselves tested.
A board full of green numbers is only worth as much as the tests behind it. So each suite is checked the hard way: a real bug is planted in the code — a dropped guard, a swapped value, a swallowed outage — and the suite is run to confirm it turns red. A planted bug that slips through is not a pass; it is the specification for the test that was missing, which then gets written. Run that loop across every suite and every planted bug has, in the end, been caught — and the few first-run misses are exactly the tests that loop added.
That is the difference between a test that anchors against reality and one that agrees with itself, made checkable. A green count above means the promised behaviours held in the last run — not that every behaviour is covered, which the board is careful never to claim.
Your fleet would have this page too.
This is not a dashboard bolted on after the fact — the heartbeat and the recorded test run are properties of how each workflow is built. Prototype to Product installs the same shape under your product: the boxes, the one place they all report to, and a status page like this one pointed at them, with your first real workflow wired end to end.