A status page that works when the product does not, and published response times

What changed

There is now a public status page. It lists the parts of the product you would recognise, anything currently wrong, and every past entry with its full history.

It is a plain file on a separate host. It does not read the product database, it runs no scripts and it makes no request back to the platform, so it loads on the day the product does not. That is the only reason to have one.

Alongside it, every support item now carries four things: a severity, an owner, a first-response target, and the target stored on the item at the moment it was raised.

What you will notice

A response time you were told is a response time that still applies later. It is written onto your item when the item is created rather than read from a page at the time you go looking, so it cannot change behind you.

The two highest severities, a product outage and a suspected data exposure, page a person at any hour. The other two go into the normal queue. If you are unsure between two levels, pick the higher one.

What has not started yet

Monitoring is not armed. Component state on the page reads unknown rather than green, and the page says so on its face. Nothing polls anything yet, because there is no external customer to poll on behalf of. The page, the protocol and the severity ladder are live now so the first real incident is handled by a procedure somebody has already exercised.

Why

Because the honest failure mode of a status page is that it is built during the first outage, by the person who should be fixing the outage. And because a response time that can be edited after you were told it is not a commitment, it is a preference.