talos linux showing as not ready after using `kexec` during decommissioning

@rothgar was testing Talos Linux on Oxide and noticed that, when decommissioning Talos Linux nodes, some nodes would remain in the Not Ready state in Omni, a Talos Linux management program, after attempting to use kexec to reboot after wiping itself.

During a troubleshooting session we collected propolis logs for an instance that failed to reboot with kexec (talos-10) and an instance that successfully rebooted with kexec disabled (talos-11).

• talos-10-oxz_propolis-server_7bcbb1b1-9f72-40e6-aecd-ca49995dc0ac.log
• talos-11-oxz_propolis-server_9468a2bd-ec9a-46da-bf1c-7f4c15706d5f.log

However, it's not yet clear what's going on here. The minimum reproduction case is still not guaranteed since maybe 3 nodes out of 25 experience this behavior when decommissioning. The following items are outstanding and would help narrow down what's happening here.

• Narrow down and document the reproduction case to reliably reproduce this issue at a higher yield. This would help in gathering guest OS logs and Oxide logs.
• Gather more guest OS information during a reproduction.
  • Output of dmesg. Can this be reliably retrieved via talosctl when the instance is decommissioning?
  • Serial console output.

Start by comparing the attached propolis logs for talos-10 and talos-11, then investigate the decommissioning path that uses kexec. Attempt to make the reproduction reliable at a higher yield and collect guest dmesg through talosctl and serial console output. Done means the failure is reproducible and the relevant guest and Oxide logs are documented.