Launch morning is three GETs and a snapshot

Launch day. The Show HN post is drafted, scheduled for 9:00 AM.

At 8:40 you run the checklist. First, a snapshot of the production container named pre-launch, so the worst case is a thirty-second restore. Then GET /api/v1/containers/{id}/proxy/groups and the merged access view per service: you read exactly which routes are public and which still answer only to your token. The admin panel isn't on the public list. Good.

9:00, the post is live. By 10:20 the front page has tripled your traffic. You keep one tab on GET /api/v1/system/resources; the box absorbs it. No autoscaler spun up, no bandwidth meter ran: a flat-rate server's launch day costs what its Tuesday costs.