Something that clicked for me while poking at my own test phone: on Android every installed app gets its own Linux user ID. The sandbox between apps is mostly the plain old Unix permission model, plus SELinux on top.
On a device you own with debugging enabled:
adb shell ls -ld /data/data/com.example.notes
# drwx------ 4 u0_a142 u0_a142 4096 ... /data/data/com.example.notes
u0_a142 is a user, created for that app. The directory is mode 700, so no other app (a different UID) can read it, and the kernel enforces that exactly like it does for two users on a desktop Linux.
This explains a lot: why apps need declared permissions to touch shared stuff, why an app cannot read another app's files without help, and why root on a phone is such a big deal (UID 0 ignores all of it).
Linux on the desktop, one human, many users in theory. Android: one human, many users in practice, each one an app.
Comparing notes with the iOS people: how does the other side do it?