Tail production stderr from your phone

Saturday afternoon. The on-call alert says the worker is restarting in a loop. You're not at a desk. The old playbook needs a laptop, a VPN, a bastion ssh, a journalctl invocation, an eye-strained scroll on a 5-inch screen.

You open the bookmark on your phone. GET /api/v1/daemon/programs/{worker_id}/logs?type=stderr&lines=200. Two hundred lines of stderr render in your browser. The error is obvious: a missing env var since the 2pm config push. You fire POST /programs/{worker_id}/stop, hit the env-var fix in your config tool, then POST /start. Status flips to RUNNING. Uptime climbs.

You did all of this from a phone, on a hike, with no app installed. You hike on. The worker keeps running.

Stderr stopped being a file you tail and became a URL you GET.