4 or 6GB depending on storage size, and these devices (ipads) run iPad apps with more aggressive memory management & not full blooded MacOS applications. Additionally, applications are usually (not always) written to support lesser iPads (with even less memory) than Pro models.
The combination of those factors are why many of us (on the other side of the argument) feel the original lack of support for Stage Manager on older Pro models was nonsensical.
I would point out that Apple lifted the limit on apps on the original Stage Manager to allow 15GB of RAM allocated to each and every one of those eight apps running simultaneously. Even on Windows machines, they have minimum RAM requirements to run anything reasonably well. Virtual memory is great and all, but even it needs real RAM for decent performance. The faster flash on M1 iPads still runs glacially when compared to real RAM. macOS has pretty good memory management but most reviewers still recommend 16GB, showing they recognize that the more real RAM there is, the better things perform. Most Mac apps do not use memory allocation remotely close to 15GB and macOS still requires 8GB at minimum, granted the OS itself has a larger memory footprint than iPadOS, but to run really well it needs 16GB or more.
Why is it unreasonable for Apple to use the metrics of apps allocating 15GB each when specifying limitations on Stage Manager? Sure, if you run eight tiny apps, you're not going to need even 4GB and no virtual memory at all, but if you're running eight apps each with 15GB in the future, and they all have to be awake the entire time, Stage Manager would screech to a halt with insufficient real RAM. They're envisioning running the equivalent of eight Photoshops at the same time, each editing a monster file, filling up that entire 15GB allocation for each app. Apple is essentially saying that the 8GB in the M1 iPads is enough, but that 4-6GB may not be sufficient. Put limitations on the lack of virtual memory support in the A12 series together with low RAM and incredibly slow flash (1/5 the performance of the M1 iPad Pro), you've got some serious obstacles to a well-performing SM. Any engineering company would use worst case scenarios when determining the minimum requirements for a feature, not the best case scenarios, hence why using a handful of apps with low RAM usage isn't all that meaningful. You're simulating a best case, not a worst case.
I've also explained earlier why Apple doesn't have full display extension on iPadOS. They don't even have that with Stage Manager. Stage Manager is not full display extension by any other name. When people think of display extension, they envision a joined desktop spanning multiple monitors where apps can even sit between two monitors, with half in one and half in the other because they make up contiguous space. iPadOS doesn't have that because it doesn't have a desktop at all. The closest comparison is macOS's LaunchPad, which looks a lot like Springboard. Note that LaunchPad is single monitor. Why is that? Because it makes no sense to have a bunch of app icons on multiple monitors even if those monitors are available. Springboard is exactly the same thing.
On top of that, iPadOS at its core is touch-centric. What's the point of a second monitor if you can't touch anything over there? This is why apps that currently support a second monitor use that only for display purposes with no live interaction. iPadOS 15 and earlier does not even supply notifications for the second monitor. It's also why Apple mirrors the main display instead of extending the "desktop". How do you extend something that doesn't exist? If you can't touch it, what's the point? Sure Apple could fill in the sides with more wallpaper to eliminate the black borders, but why? It would still look ugly with nothing in them, with icons clustered in the middle and the edges empty. So they don't bother. It wouldn't be hard to fill the sides, but it doesn't make any sense to do so. The only way Apple will do that is if a future iPad moves to 16:9 or 16:10 aspect ratio. It'll still mirror.
You'll also notice Stage Manager differs from the version on the Mac. You cannot freely drag apps between monitors, even in the full blown iPad version. The reason is because there is no desktop. You have to either use those three dots or use the Globe-Control-Backslash key combo to move the frontmost app to the other monitor. Stage Manager consists of two entirely self-contained packages of four apps, and never the twain shall meet. They have their own docks and their own side app panels. Even with just a self-contained Stage Manager on the second monitor, it is so buggy that they had to pull the feature for later release while SM on the iPad itself works pretty well. You might wonder why is one stable and the other unstable since they look exactly the same. Non-engineers make the fundamental mistake that just because something looks the same that it is. This is another reason comparisons to 2006 desktops is non-sequitur. Bottom line is that secondary monitors make no sense on a touch-centric device, so iPadOS would need major revisions to use one, along with an entire conceptual shift in creating an actual desktop environment where none currently exists. Opening up a secondary monitor for generic use is opening up a can of worms Apple doesn't want to deal with. If they do eventually move to a desktop setup, expect that to take many more years because it is essentially a re-writing of major parts of iPadOS rather than the tacked on module that SM is. Samsung does it with DeX for two reasons. They also completely isolate DeX from its core Android UI so they avoid any fundamental changes and they need to put lipstick on a pig. Android lacks tablet apps and DeX is there to make ugly phone apps look less ugly. It's not very useful, though. I own two Galaxy Tabs (S7+ and S8 Ultra) and DeX is pointless.
If I were a programmer at Apple, I would have created Stage Manager as side code, running separate from core iPadOS to avoid having to change fundamental aspects of the OS, and from how SM runs and how there's zero interaction between the two monitors, I think they did exactly that. What you're asking for would require a fundamental re-writing of iPadOS, and I don't think Apple is willing to do that to satisfy the teeny tiny minority who actually use a second monitor.