Hey @mathroc thanks for the question! This is something we’ve been thinking about internally for a while and are definitely planning to go ahead with. You’ve hit the nail on the head with the considerations around releases, there are a lot of caveats and a lot of things to consider UX-wise with it. Not to say this is how it will be implemented since things can change, but we were thinking that within the settings of a fleet you’d be able to toggle additional architectures, that would then allow new releases to be built that contain the images for multiple architectures as you suggest. A release would not be marked successful until it contained images for all of the architectures specified by the fleet. Of course, as you point out, when you add an architecture to a fleet then you’d have to create a new release in order for it to be supported.
At the moment though the two options you’re considering are the way to go - depending on the devices you want to use, choosing something like a Pi Zero for your fleet device type will give you a lot of options. This is what I normally do when I want to run an application across a mix of different Arm based devices. With regard to the two fleets, you’ll note this is what some of the fleet owners on balenaHub do in order to get around this problem, and we also did the same when we launched the Fold for Covid open fleet.
I’ve added your request to our tracker alongside @maggie0002 who was also asking for this same feature.