Excessively long duration reported for "Queued" spans

[seh]

In a recent trace published for a workflow run using this action as a final job within the same containing workflow (per #14 (comment)), one of the trace spans has the name "Queued" and a reported duration of 18446743100709.55 milliseconds, or 5,124,095 hours and eighteen minutes. The delay was never apparent while the workflow was running.

Is this action misinterpreting a duration reported by GitHub's API? I see that there's special handling of this span in this action's code:

otel-cicd-action/src/tracing/job.ts

Lines 139 to 140 in a335d59

	const queueSpan = tracer.startSpan("Queued", { startTime }, queueCtx);
	queueSpan.end(new Date(firstJob.started_at));

[corentinmusard]

Same issue as inception-health#201.

Currently, Queued span is set with the following [startTime, endTime] interval:

• startTime is new Date(run_started_at || created_at) from getWorkflowRun endpoint.
• endTime is new Date(jobs[0].started_at) from listJobsForWorkflowRun endpoint.

Is your workflow run public? We could check the values I mentionned above from these two endpoints to see if it's github's API fault or not.

[seh]

Unfortunately, no, it's not public. I haven't noticed this happening again since I filed this issue, but I don't inspect all of the traces. Whatever the cause, it's intermittent and occurs only very rarely for us.