It seems like a flake as it happened twice and a later run didn't fail with the same error.

For the most recent failure, the details are visible here:

https://luci-milo.appspot.com/buildbot/chromeos/tricky-chrome-pfq/2780

The logs aren't very helpful, as a number is used instead of a name.

    01/05 01:52:01.187 DEBUG|          parallel:0030|   File "/usr/local/autotest/tests/desktopui_ExitOnSupervisedUserCrash/desktopui_ExitOnSupervisedUserCrash.py", line 43, in run_once
    01/05 01:52:01.187 DEBUG|          parallel:0030|     raise error.TestError('Failed to retrieve session state: ', e)
    01/05 01:52:01.187 DEBUG|          parallel:0030| TestError: ('Failed to retrieve session state: ', DBusException('The name :1.114 was not provided by any .service files',))
    01/05 01:52:01.228 INFO |               job:0206| 		ERROR	desktopui_ExitOnSupervisedUserCrash	desktopui_ExitOnSupervisedUserCrash	timestamp=1483609921	localtime=Jan 05 01:52:01	('Failed to retrieve session state: ', DBusException('The name :1.114 was not provided by any .service files',))
    01/05 01:52:01.235 INFO |               job:0206| 	END ERROR	desktopui_ExitOnSupervisedUserCrash	desktopui_ExitOnSupervisedUserCrash	timestamp=1483609921	localtime=Jan 05 01:52:01

The test code instantiates a session_manager.SessionSignalListener(), and is trying to wait for a 'stopped' signal, and the error is raised while waiting for it.

I would have cc'd cmasone@ as the author of the session_manager.py code used, but he no longer appears to be involved.

Filing the bug, and keeping it open/unassigned for now, to see if we see more reproductions.