...except in virtually any case where you'd run something untrusted there you'd use Flatpak or something similar where what you wrote doesn't apply.
I think the important distinction is _everything_ should be considered untrusted because even trustworthy software can become malicious. For example, the XZ Utils backdoor[0].
On Android, everything I run is subject to the permission model and sandboxed. That is not the case on Linux.
1. By being baked into the OS itself, which is unavoidable since the OS is the thing providing the sandboxing + security model. It still massively reduces the attack surface.
2. By being run through the android debug bridge, which is far from normal and something users have to explicitly enable. Leaving you the option to shoot yourself in the foot in an opt-in manner 99.9% of users will never touch isn't the same as Linux where foot-shooting is the default.
If you want to confine yourself in a sandbox, feel free to do it. The past decades have demonstrated that it's only necessary for some specific threat models.
I want to confine apps in a sandbox. Android has that, Linux... well not really. I mean "it's possible", but it's not integrated like in Android.
You can configure your flatpak app so that it will have permission to read microphone in the background or have full access to the disk. Many flatpaks of real apps request dangerous permissions that users have been conditioned to ignore. For example Blender is such an app which has full disk access and background microphone access, and I'm sure many people have installed that. This is unlike Android where these are locked down for every app.
The world isn't black and white. Most reasons why Android apps are being so heavily locked down don't apply to Blender. As a user, I'm not interested in Android-bis - if I were, I would just use Android after all. Nevertheless, things like Flatpak give me, the user, the power over application's permissions and I can take them away (or give more) in a few taps at will. The defaults being tuned for different use cases and threat models are not "being decades out of date", especially when you could already use the existing tooling to replicate other models - regardless of whether you happen to like these defaults or whether they fit your specific use case.
This is not possible on a device following the Android security model. Permissioned features should always be implemented using proper security mechanisms like permissions.
>don't apply to Blender
You say that until the SuperRetopologyTools5000.py addon you try out infects your system.
>Flatpak give me, the user, the power over application's permission
Most people are not going to bother with this. It's important for the defaults to be secure. People shouldn't have to opt in to a secure experience, and doing so shouldn't break the program.
Yes, that's the user experience on such devices. "This is not possible on a device following the Android security model, either bypass it or use another device".