As someone who has experimented with open-source serverless/Function-as-a-service before (like OpenFaaS and knative), I would say that serverless sucks if you want to use it for interactive, burst-period and time-of-day-sensitive stuff.
You better treat each function to run dedicated stuff for one specific thing, so do one thing and do its best, and you need to make sure that it is idempotent and not fatal to repeat -- trivial things like sending e-mails, making snapshots off a website using Selenium/Puppeteer, image/video conversion, all fall into this category. At that Eureka moment I realized why AWS called it Lambda.
This idea that serverless is "just" for scalability is a misconception and pessimism about the future of computing. What serverless should really mean is that you focus on your code and you don't configure your digital ocean, the node.js runtime, the observability, the security, the reliability, and all this shit.
If you couldn't just click on an icon to open a web browser on your laptop, no one would be using the internet. That's the same with cloud computing. The amount of people who use the cloud is <1% of what it could be.
So go ahead and pay $8/month for the hosting, then pay with your own invaluable time for the rest.
You better treat each function to run dedicated stuff for one specific thing, so do one thing and do its best, and you need to make sure that it is idempotent and not fatal to repeat -- trivial things like sending e-mails, making snapshots off a website using Selenium/Puppeteer, image/video conversion, all fall into this category. At that Eureka moment I realized why AWS called it Lambda.