Comment 5 for bug 1891076

Is this a case where we were looking in the wrong location for the logs? Eg machine-1 is failing to start unit-foo-1 but we were looking for unit-foo-1.log but it doesn't exist yet, and we didn't check the machine-1.log.

We certainly could record errors of this sort against the unit that we are failing to start (so we have more detailed information about the failure than 'allocating').