Thank you! Simple is indeed one of the goals.
Currently, a particular site exists as static files on one specific server, with some failover mechanisms in place. So mainly availability is achieved by having enough server resources and having a minimum amount of moving parts :)
The tricky thing to solve for with HA is TLS termination. It is probably the most complicated thing (and for sure the most computationally intense thing) a static HTTP server does. I haven't done any testing, but my hypothesis is that TLS termination is what will start failing first. Barring DNS-based tactics (mitigation, not cure, in my opinion) you just have to accept that this is your single point of failure.
But a great question! Keeping our sites online is of course something we're working on all the time.
And yes, it was all fine before the cloud! Alas, a single server was (much?) more stable in the early days than a single instance/node/pod is today...