3 Deployment Fails That Made Me Quit Oracle Cloud Forever

This is Part 2. Part 1 covered 3 bugs I hit while building the video merging engine. This one is about what happened when I tried to put it online. The Plan That Looked Too Good In Part 1 I built Merge Video — a service that merges dozens of video files using ffmpeg and uploads the result to YouTube. I film long strategy sessions — 6 to 8 hours each — and end up with 30–50 clips per session that need to be merged into one video. It worked on my machine. Now I needed to deploy the Telegram bot so it runs 24/7 without my laptop being open. Two requirements: Free — at the prototype stage, free tier makes the most sense. Cloud providers offer generous free tiers specifically for this use case — validating ideas before committing budget. Once there's real traffic and real users, paying for infrastructure is a no-brainer. But spending money before product-market fit is just burning runway Persistent — a long-polling bot needs to stay alive, not sleep after 15 minutes of inactivity like Rende