SCENE_01
Your junior asks to own the service
It's Tuesday afternoon. You're in a review comment thread when she messages: she wants to take over the on-call rotation and the deployment process for the cache layer—the one you built, the one you still touch every sprint. Your chest tightens before your mind catches up. She's ready. She's more than ready. And that's the problem.
SCENE_02
What it means when you say no
If she runs it alone and nothing breaks, the story rewrites itself: the service didn't actually need the architect who designed it. The careful abstractions, the defensive checks you added—were they solving real problems or solving the problem of staying useful? You picture your manager six months out, no incidents, no heroic fixes, just steady operation. The thought lands: I'm not indispensable. I'm overhead. So you find reasons to stay involved. You add another abstraction layer that only you fully understand. You keep the deploy script on your machine. You schedule reviews instead of publishing the runbook. She stops asking.
SCENE_03
The reading that changes what happens next
Here's what actually happened: she was ready. The service works because it was built well. That's not a loss—that's the job. The next problem you solve is where you're senior, not the last system you guard. If you stay threaded into every decision, you've made yourself into the bus factor you warned about in every architecture meeting. That's not loyalty. That's a design flaw. Write the runbook. Let her deploy. Watch what you actually become when you're not busy proving necessity.