The agent necessarily runs on the same machine as the job, which gives the job itself control over the clock used to provide the time_remote time stored for each job event record. In addition to providing this wall time, we should have the agent provide the monotonic time it read for each event: this time is immune to NTP or other clock adjustments by the job itself, and would at least provide a valuable debugging aid.
The agent necessarily runs on the same machine as the job, which gives the job itself control over the clock used to provide the
time_remotetime stored for each job event record. In addition to providing this wall time, we should have the agent provide the monotonic time it read for each event: this time is immune to NTP or other clock adjustments by the job itself, and would at least provide a valuable debugging aid.