Morning friend ☕️

Episode 11 of 13.

Last week, Thandi moved Michael ahead of the system. This week, the system breaks in front of everyone.

And when something fails publicly, the first decision doesn't fix anything. It decides what spreads.

Grab your coffee. Let's chat.

At 9:17am, the dashboard turns red.

Not fully.

Just enough to remove certainty.

The war room fills in minutes.

Werner at the board. Michael on his laptop. Two engineers already talking over each other. A business lead standing before anyone's asked him to sit.

"We escalate now."

Thandi looks at the screen.

No pattern yet. Just noise.

"Not yet."

Silence.

"We escalate when we have signal."

Someone reaches for the phone.

"Leave it."

The hand stops.

She walks to the board. Crosses out GO LIVE.

Writes CODE RED.

Under it:

Incident Lead: Thandi
Update: every 15 min
Escalation: pending signal

She watches it land.

If this goes wrong, her name is already on it.

She nods once.

That's the decision.

For three hours, they look in the wrong place.

Gateway logs: clean. API calls: correct. Network: fine.

The failures look random because from the application layer, they are.

At 13:41, a finance analyst joins the call.

"We've got a reconciliation problem. Transactions are showing approved. But balances aren't moving correctly."

Werner looks up.

Michael turns.

That's not where anyone was looking.

The investigation shifts to the database layer.

Forty minutes later, someone asks the question that breaks it open.

"What IOPS did we provision on the storage? And what's the ceiling on our instance?"

Two different numbers.

Two separate decisions. Made at different times. Never checked against each other.

The storage could handle the load. The instance had a hard ceiling it couldn't cross.

Under normal traffic, they'd never come close. On launch day, with real users hitting the payment database all at once, they pushed past it. IO requests queued. Queries slowed. Transactions timed out mid-write.

The problem was already there.
They just hadn’t met it yet.

The fix lands at 8:42pm.

Not instantly.

Emergency change approval. Staging validation. The AWS modification itself. A brief failover where the payment database isn't there at all.

Then it is.

Eleven hours. Twenty-five minutes.

At 9:03pm, the dashboard turns green.

The room erupts in the way only people who've seen the edge of something and survived can laugh.

Thandi doesn't.

Her coffee from 9:00am is still on the desk. Cold. Untouched.

She opens her laptop and starts the incident report.

The reconciliation job runs quietly in the background. Three days of automated work. Some transactions will take weeks to resolve.

She doesn't mention that.

The room doesn't need to know tonight.

The next standup runs differently.

Nobody guesses out loud. Nobody fills silence with a status that isn't ready. Werner flags a dependency before Thandi asks.

The team didn't become better engineers overnight.

The room just stopped working against them.

The IOPS mismatch was always there.

It was there before launch day. Before the war room. Before the dashboard turned red.

What changed at 9:19 wasn't the technical problem.

It was the conditions the team worked inside.

Panic doesn't just feel bad. It degrades thinking. It shortcuts diagnosis. It makes noise look like signal and speed look like progress.

Thandi didn't fix the system.

She protected the conditions that allowed the team to fix it themselves.

That's a different thing.

And it only works if someone is willing to hold the room before they know how it ends.

When a system breaks, most managers ask:

How do I fix this fast?

A harder question arrives in the first two minutes.

What am I allowing to spread right now?

Because fear moves faster than any fix.

And whatever the room catches from you early becomes the operating system for everything that follows.

Crisis leadership isn't proved by how fast you fix things.

It's proved by what you protect while they're still broken.

Until next week, Vaugan ☕️