I’m guessing you figured this out by reading the sources. In which case: Thank you for putting in the work to summarise it for us. It’s pretty interesting. WayDroid makes a lot more sense now that you’ve explained it.
For some reason, I had reached the conclusion they were running a Wayland server inside the container, rather than a client, so that apps on the host would have to connect into the container to display a GUI! Which is why I said “That sounds pretty drastic”!
Armed with the knowledge that both use LXC, I realise I also misunderstood the following statement from the Linux Smartphones article:
Unlike a typical emulator, WayDroid uses the same kernel as the host operating system, allowing the Android apps to interact with a device’s hardware almost as if they were native apps
On first reading I took it to be a comparison of WayDroid against Anbox, with Anbox being considered a “typical emulator” for the purposes of the comparison. But clearly this statement actually applies to both WayDroid and Anbox, compared against unspecified other “typical emulators”, which would include virtual machines running Android.
This is what I get for skim-reading.