RetroN 5 is officially a thing, apparently

[Ed Oscuro]

Hyperkin hasn't said anything and I doubt they will until it's released and people take a look inside. However, my guess is that the Retron5 will be essentially a computer that has cartridge support. There are several clues that lead me to this conclusion:

I agree it's likely to have either a really cheap x86 or more likely an ARM CPU inside.

720p video output. Analog RGB signals, even when converted into digital progressive scan, can produce a maximum 240p.

Analog signals can be interlaced or progressive; I don't know what you mean by "converting" here. Most televisions with support for 480p progressive signal used YPbPr (aka "component") video.
The PC VGA monitor is one example of a type of analog RGB connection which is usually much higher than 240p; I'm using this analog connection right now to type out this response - on a 1920x1200 monitor! (This is about as high as you'd want to go with analog RGB because the signal will eventually start to reach its bandwidth and clarity limits).

720p is an HD resolution that simply not possible with the original hardware of any of the systems supported.

I don't know what you mean by "simply not possible." You can get an (analog) 720p signal from an upscaler, and without modifying the system.

If you wanted an old-fashioned console to output 720p, it would be a relatively simple engineering task. You could either do it by converting the analog RGB signal (or whatever you have to work with, actually; nothing says you can't work from composite or S-Video) or you could do it (like somebody has recently been working on with the NES) by intercepting certain signals from "original" NES hardware (the CPU etc.), or redesigning the board slightly (same thing) so that instead of sending data to analog RGB encoding, you send the data to be encoded in a digital system. Nothing magical or unusual about it. Again, you can use an external upscaler or converter box to change signals to pretty much any format you need, depending on what you're starting with and what you're going to. It's just more elegant and probably much faster to do what I am sure the RetroN is doing, which is generating the data which will make up an image internally (whether by emulation or with some close-to-original hardware makes no difference, seriously) and then generating that as a digital image, rather than an analog one.

Savestates, button remapping, speed settings, audio interpolation, built in cheats, etc. These are all standard emulation options only a computer offers.

Again those are actually things that could be done without emulation. I've got controllers that can be remapped, I've got a cartridge-taking MSX computer with a CPU speed slider, we've all got Game Genies, and for the most likely of the features - savestates - it's actually not impossible to dump or intercept the state of a CPU. It's just not a feature that works very well (if at all) with the cheap CPUs found in these systems. Nothing theoretical says you couldn't design a new CPU that has it.

Future firmware updates. Firmware means an OS, means computer.

You don't need an OS to have firmware. Furthermore, the firmware could just be for tweaking peripheral functions or video output.

"Our aim is 100% compatibility". The only way to achieve that through clone hardware is through emulation, which means a computer.

Emulation is actually a good way to screw things up, and emulation is very rarely "cycle accurate" to the original hardware (MAME isn't, for example), and that will likely be needed to really get some SNES effects right. And if you look at the current builds of higan (formerly bsnes) you'll see that requires a lot more CPU horsepower than this thing is packing. Another approach (which is a kind of emulation) would be the use of a FPGA but that's even less likely to be successful, especially since the emulation community has focused on emulation software, rather than implementing things in FPGAs (and FPGA implementation details can and will differ significantly from device to device). And then there is another, much more obvious route you seem to deny is possible - using original components or close analogues (if you recall, the SNES CPU was intended in part to allow NES gameplay on the system, and the Genesis 3 also has some differences from the original Genesis which affect a very few games). Building a system with fully digital output, without tons of signals from different chips interfering with each other, and without terrible expense, doubtlessly ruled this approach out. But I don't see what on earth stops a clone manufacturer from reverse engineering the very few proprietary or protected things in most of these machines, and buying off-the-shelf components for most of the rest.

Anyway, the point of this post is to point out that technical ability and cost are what will certainly cause this to run an emulator (rather than even an FPGA). However there is, again, nothing which theoretically forbids it. It's induction based on what is probable, not deduction based on what is possible.

And what is possible is probably a lot nicer than this thing will end up being. However, it's already very expensive for people to have little batches of just small bare PCBs made, to populate with off-the-shelf components, to say nothing of trying to make new designs, especially if they require a CPU fabrication plant to make new stuff (very expensive) or have hybrid components. However it's certainly possible to do some of the stuff mentioned, like have a digital signal with some user-selectable memory sections displayed in the corners, or a built-in cheat engine to change data (just halt the CPU first, easy even on cheap CPUs), and to do some of the other things mentioned.

diehard classic gamers [...] already have the consoles, they're more durable than any cheap 21st century Asian clone will be, you're guaranteed of 100% accuracy, compatibility is guaranteed,

Well, not exactly - firstly, old things can die. Ever heard of caps? I myself don't have a soldering iron; I have to ship a few things out soon to get them fixed. And other, more troublesome things can and do go wrong - RAMs often go bad. Batteries leak acid and destroy PCBs and other components. Components get fried by power surges. Even getting that "classic look" takes a classic tube TV, I find, and good ones are dying all the time, and not being replaced. The old consoles often can't physically connect to a modern TV without a fair amount of technological wizardry connecting components.

100% accuracy and compatibility - well, the Genesis 3 isn't 100% compatible with the Genesis game library. (It is a better system than I used to think though, and the broken games don't really matter.) Accuracy could even go bad if an oscillator goes nuts or some other component fades in an unusual way. The CPU and other burned-on components (like mask ROMs, not always used in cartridges though) should last and perform accurately a very long time, but they do rely on a bunch of other unusual stuff to make the system go.

We're in a situation where there is a constant attrition of older, original consoles. Really, I think the only thing that will save some of these libraries will be the eventual release of a console clone, either by the original manufacturer or a third party, with modern-day output features (likely to remain digital signals for the foreseeable future). Other than that, I don't know, I guess I sound like a survivalist now - stock those systems while you still can!