In Scott's local instances he gets "systemctl is-system-running" == "starting" with
JOB UNIT TYPE STATE
2 dev-sda2.device start running
So there is some impedance mismatch between is-system-running and Type=idle -- apparently the idle unit already starts while the above sda2.device is still running (or, perhaps, it flips *back* to starting while it is already running).
In both Scott's VM and local LXD I never get dev-sdaX.device as "running", only as "activating tentative" (and that does not time out); consequently is-system-running does not say "starting" after reaching multi-user.target, and invoke-rc.d DTRT.
In Scott's local instances he gets "systemctl is-system-running" == "starting" with
JOB UNIT TYPE STATE
2 dev-sda2.device start running
So there is some impedance mismatch between is-system-running and Type=idle -- apparently the idle unit already starts while the above sda2.device is still running (or, perhaps, it flips *back* to starting while it is already running).
In both Scott's VM and local LXD I never get dev-sdaX.device as "running", only as "activating tentative" (and that does not time out); consequently is-system-running does not say "starting" after reaching multi-user.target, and invoke-rc.d DTRT.
Scott's debug journal: http:// paste.ubuntu. com/23153994/
So currently I lack a way of reproducing that "running" state.