Comment # 4 on bug 1228438 from Fabian Vogt

(In reply to Franck Bui from comment #3)
> Just providing some updates based on my understanding of the test after
> Fabian did some investigation (thanks a lot Fabian).
> 
> Apparently `wsl /bin/bash -c "systemctl is-systemd-running` creates an
> environment (container) based on the host environment before executing bash
> in the container.
> 
> More specifically, the hybrid cgroup hierarchy of the host is reused by the
> container. However since commit f2512de82dc91cfb742a4f4df934bdb4fcad482d,
> systemd makes systems based on the legacy cgroup harder to boot by incurring
> a delay of 30 sec to the boot process.

I filed a new report at WSL: https://github.com/microsoft/WSL/issues/11857

Maybe it's possible to somehow work around this downstream by ignoring the
inherited cgroup hierarchy and just setting up a v2 one?

> This probably explains why the command `wsl /bin/bash -c "systemctl
> is-systemd-running` is failing to contact PID1 in the container. Sleeping
> for 30 sec before executing the command should work around the issue.
> 
> That said it's not clear to me how this is supposed to work reliably even
> with v255, ie how `wsl` figures out when it's safe to contact PID1.

I have no idea. That's completely in opaque, proprietary and undocumented code.

> PS: Fabian, please correct me if anything I wrote is incorrect.

All correct.