Fixing yt-dlp in Docker: n-challenge EJS scripts, Deno 2.x, and the player_client=ios cookie trap

If you run a video processing worker that downloads YouTube content with yt-dlp, you've probably hit the n-challenge wall: downloads stall or fail with a Sign in to confirm you're not a bot error, or the n-parameter isn't being decoded correctly, leaving you with throttled or broken streams. After a few rounds of debugging our Railway-hosted worker, here's everything we fixed and why. Background: what is the n-challenge? YouTube applies a per-request throttling parameter called n to video URLs. yt-dlp has to solve a JavaScript challenge (embedded in the YouTube player) to decode this parameter into a valid token. If the challenge isn't solved, requests are throttled to ~50 KB/s or rejected outright. The solver is written in JavaScript, and yt-dlp ships with multiple backends for running it: a pure-Python fallback, and native JS runtimes (Node.js, Deno, PhantomJS). The JS path is dramatically faster and more reliable — but only if the runtime is actually available and the yt-dlp install