In the gnawty-release build at http://stainless/browse/chromeos-autotest-results/225399103-chromeos-test/, the power.Reboot test failed due to the reboot command timing out:

2018/08/09 09:15:38 Started test power.Reboot
2018/08/09 09:15:38 [09:15:38.642] Rebooting DUT
2018/08/09 09:20:38 [09:20:38.643] Error at reboot.go:27: Failed to reboot DUT: context deadline exceeded

The messages log shows that the system actually rebooted, though:

...
2018-08-09T09:15:39.375154-07:00 NOTICE pre-shutdown[12372]: Shutting down for reboot: not-via-powerd
...
2018-08-09T09:15:39.397593-07:00 ERR wpa_supplicant[587]: nl80211: Failed to set IPv4 unicast in multicast filter
2018-08-09T09:15:45.216842-07:00 INFO kernel: [    0.000000] Initializing cgroup subsys cpuset
2018-08-09T09:15:45.216914-07:00 INFO kernel: [    0.000000] Initializing cgroup subsys cpu
2018-08-09T09:15:45.216918-07:00 INFO kernel: [    0.000000] Initializing cgroup subsys cpuacct
2018-08-09T09:15:45.216920-07:00 NOTICE kernel: [    0.000000] Linux version 4.4.146-14675-gf9e8bb4004dd (chrome-bot@swarm-cros-464) (Chromium OS 7.0_pre331547_p20180529-r9 clang version 7.0.0 (/var/cache/chromeos-cache/distfiles/host/egit-src/clang.git 1084e053422cbbaaa6e06a972aac210489e7a0ad) (/var/cache/chromeos-cache/distfiles/host/egit-src/llvm.git 
...

I think that my code in power.Reboot to run the "reboot" command is buggy. It attempts to run a subshell that sleeps before rebooting (so the SSH exec request will finish successfully), but it actually blocks until the reboot command returns, resulting in a race where the command will appear to hang if the system reboots too quickly.