Cilium Feature Proposal

When upgrading our Kubernetes workers from Ubuntu 20.04 -> 22.04 we encountered the issue described in #10645, since Ubuntu 22.04 ships with systemd 249.

The workaround is described in the System Requirements page, but I only managed to find it there because I was looking for it (I ran into this issue a few weeks ago trying to run Cilium on my local Arch workstation).

Given that the failure mode is really weird (host networking pods work fine, networked pods work fine when talking to same node), perhaps it would be worth checking the sysctl setting during cilium-agent startup, and fail to boot if it's not correctly set?

With Ubuntu 22.04.1 (the "real" LTS release, as everybody knows) around the corner, I suspect that many Cilium users are going to encounter this issue in the short term. Adding a check could save them (and folks responsible for Cilium issue triage!) a lot of time :)