For what it’s worth, most of what I’ve seen people having trouble with, as far as changes and updates to Anvil goes, has been with the beta editor. The classic editor has been very solid in my experience (basically the Anvil incarnation of an ‘LTS’ version), and so it’s currently my only choice for production. When the Anvil team gets the beta editor stable, then you may consider using it. I think the biggest benefit of the beta editor is being able to install your own choice of libraries (or new versions of libraries) in your server account. If your needs are satisfied by default libraries installed with the classic editor environment, then you should be able to use it and have a much more reliable experience than you may have seen people discussing about the beta editor.
If I was in your position, and basically zero downtime was required, I’d consider using Anvil’s hosted service for development work, and run the open source server on AWS for production. That would give you all the productivity benefits that Anvil provides (which I think are nearly impossible to match with other tool chains), and the safety of controlling the production environment as you see fit. In my opinion, the trade off that comes from having to run your own server on AWS doesn’t deteriorate the productivity benefits of using Anvil, enough to make me want to use another tool chain all together. And the truth is: if you hope to control your environment enough to maintain zero downtime, you’re likely going to have to do at least as much work supporting any other tool chain and production build setup as you would supporting Anvil open source server - with none of the potential development productivity benefits that Anvil enables.
In my experience, the more complexity you add to your entire tool chain, the more likely you are to have problems of all sorts. Anvil’s purpose, and the reason it’s currently the most reasonable choice for me, comes from that fact that it dramatically reduces so much complexity in all phases of the development and deployment process. That not only makes the development routine a pleasure, and reduces development time, reduces costs, etc. - it also reduces the total number of potential problems that are encountered in every phase of development and production on any project. You’re going to run into problems with any development and deployment toolchain you choose. You’re going to have to put time, money, and work into maintaining the environments and tools you choose to build your product. Maintaining Anvil open source will take some more work than simply hosting on Anvil’s servers, but that’s a choice you can choose to make. There is a community of Anvil users relying on Anvil’s hosted plans for commercial clientele, and so far that service has been reliable for those users’ needs (I would expect their requirements are just as demanding as yours are). Perhaps that’s a better question to ask - has any of that community experienced a failure in the Anvil hosting environment that has led to significant disruption in business. I haven’t heard of that being the case yet (except from users who just don’t know how to use Anvil). You’re going to be relying on third party services in some way or another, unless you own and manage a data center and all your own infrastructure, in a way that’s more robust and stable than, for example, AWS, and have a team of engineers more capable than those in charge of maintaining Anvil. So far, my experience has been that’s not likely going to be the case, unless you’re managing a very large team and have a tremendous budget.
…just my few little cents to add to the pile of perspectives. Good luck!