The night an entire cloud region went down under us, I watched something that no on call policy can produce. Engineers joining the incident bridge who were not on call, had not been paged, and had every right to be asleep.
Nobody summoned them. The recovery ran for thirty hours, and the people who owned those systems treated the outage the way you treat a fire in your own house. Not because a rota said so, but because somewhere along the way, the production had become theirs.
The lazy explanation is great work ethic, which is a way of crediting the individuals so you do not have to examine the system. I have managed long enough to know better. Work ethic is real, but it does not explain why the same engineers who show up at 2 AM on one team quietly ghost their pagers on another. The variable is not the people.
The 2 AM behaviour was decided months earlier, at 2 PM, when nothing was burning.
What actually earns it
It is not perks and it is not speeches. In my experience it is four specific, boring, daily things.
Their leave is sacred. Nobody gets called on vacation. We cross train specifically so that is true, and the team settles its own leave calendar like adults. An engineer whose rest is protected does not ration their energy defensively. They give it freely, because they have learned it will not be stolen.
Credit flows down, in public. When the project lands, their names go in the email to leadership, and they present their own work upward. My name appears when something breaks. An engineer who gets the credit for a system starts to feel what is literally true. It is their system.
They are in the decisions. Tooling, process, rota design. The team decides, with full context, because they live inside the consequences. You cannot hand someone the outcomes and withhold the choices and then act surprised that they feel like a bystander.
They are never resources. Not in meetings, not in planning docs, not in my head. The word matters because the word leaks into behaviour. Resources get allocated. People get respected. Only one of those shows up at 2 AM.
None of this is charity. It is a chain with no skippable links. Respect makes the production theirs, ownership follows, and once ownership exists, accountability stops being a conversation you ever need to have. I have never once had to ask an owner whether they will handle something. The question does not survive contact with genuine ownership.
You can buy attendance. You can buy hours, and with enough money you can even buy compliance. The feeling that this system is mine is not for sale. It is reciprocated or it is absent.
The boundary that keeps this honest
Here is where I have to be careful, because this philosophy has an evil twin.
If a manager reads "my people show up at 2 AM" and hears free capacity, they have turned respect into exploitation with better branding. The willingness is a gift. The moment you treat it as a budget line, planning around it, expecting it, spending it, you are strip mining the very trust that created it, and it does not grow back at the rate you are extracting it.
So the same philosophy comes with obligations that run in my direction. I answer first at 2 AM. The leader eats the first hour before anyone else is asked to. Incident nights are followed by comp time off without anyone having to request it. And I watch for the pattern where the same engineer is the hero twice in a row, because the person who always shows up is one quarter away from being the person who used to work here.
Protect the gift and it compounds. Spend it and it is gone.
The morning after
Somewhere past hour thirty, the estate was back, the region was healing, and everyone finally got some well deserved rest.
The next morning I did not assign the postmortem. Drafts were already appearing in the channel, from the people who had been on the bridge, unprompted, because owners do not wait to be asked what happened to their own systems.
That is the whole thing, really. It was never about the night.
It was about all the days before it.
Top comments (0)