At that point, you have to intervene, troubleshoot, and fix the problem.
But if you had gone the Functions-as-a-Service route, you'd avoid such occurrences altogether. For some people, that's an appealing trade-off.
I feel like I have the skills to handle that, or to create redundancy if necessary for reliability, so that doesn't bother me, but I can see the appeal of ignoring all of that.
I still like the idea that I'm not trusting a "black box" to run my code. I know how it's set up and what it's doing. And if I add replication or load balancing, I also understand what's actually happening there. But you're right, once you get to that level, you're out of the realm of "simple".
Maybe I can still hide behind my earlier comment "unless you need the scalability..." to say that covered your caveats, but I'll admit I wasn't explicit about the considerations that guide the optimal use cases.
I don't spend much time "maintaining" my server for this kind of setup. Once it's setup I basically ignore it, accept for uploading new code.
I suppose I do occassionally run the system update command. I don't run any other kinds of PHP applications (WordPress, Drupal, etc) on an environment that I use this way, so I don't feel the constant need to check for security vulnerabilities.