How this works. Every five minutes the server requests each project through its
public address, exactly as you would: DNS, the TLS certificate, nginx and the app itself all
have to answer for a check to pass. A redirect to a sign-in page, or a password prompt, counts
as up: the app is running and guarding itself.
The two NTP Pool server rows are the exception: the time server is not HTTP, so each
of those rows is a real NTP exchange over UDP (mode 3 in, mode 4 out), once for IPv4 and once
for IPv6. They pass only when the box answers with a synchronised clock — a stratum of
1–15 and a leap indicator that is not "unsynchronised" — so a server that replies with a bad
clock fails here just as it would at the pool's own monitors.
The Server row counts the checks themselves. If the whole machine is down, no checks
run, and the gap shows up there. An independent monitor outside this server watches for that
too, so an outage gets noticed even when this page can't report it.
No cookies, nothing stored, no third-party anything. Also on this host: server stats and visitors. Raw data: status.json.