Whenever I read posts like this with millions of requests per small time unit, fighting over dollars. I realize I work in a completely different realm. I have barely 60 concurrent users but these users are paying top dollar for the privilege. Can't you make the revenue per request higher so this becomes a none issue ?
Going by your bio one could think this is business oriented. Sorry if I missed the mark. Anyway, I doubt anyone is going to run the numbers for you. I would give this a real life test, even if it was just simulated traffic. Hetzner has been a huge cost saving for me.
Even though I use systemd to manage my primary node app, I started to use pm2 for new projects. It was just easier, better monitoring, cluster option, many easy to configure restart options. I don't think the article makes a good enough case to not use pm2.
I've been playing with Remix.js a bit and come to similar conclusions. I'm not exited to go all in on route actions. I found myself using socket.io with TS interfaces to define the API. I guess pretty similar to how trpc works albeit with a bit more boilerplate. Currently looking into Vike (vite-plugin-ssr) to cut down on meta dependencies.
Samsung doesn't allow security updates while roaming (requires wifi). Another artificial restriction without much consideration for longer term travelers.
Are you interested in knowing how I did it?
Next in the newsletter I will publish a post about it, I made a system that uses Puppeteer to mount a headless browser and a frontend deployed on netlify, I publish everything to a database and automatically generate new posts on HackerNews leveraging ChatGPT