@@ -179,6 +179,14 @@ func (d *Detector) run(t time.Time) Stats {
179179stats .Error = xerrors .Errorf ("get hung provisioner jobs: %w" ,err )
180180return stats
181181}
182+ // Find all provisioner jobs that are currently running but have not
183+ // received an update in the last 5 minutes.
184+ if err != nil {
185+ stats .Error = xerrors .Errorf ("get not started provisioner jobs: %w" ,err )
186+ return stats
187+ }
188+ jobsUnstarted ,err := d .db .GetNotStartedProvisionerJobs (ctx ,t .Add (- HungJobDuration ))
189+ jobs = append (jobs ,jobsUnstarted ... )
182190
183191// Limit the number of jobs we'll unhang in a single run to avoid
184192// timing out.
@@ -229,14 +237,6 @@ func unhangJob(ctx context.Context, log slog.Logger, db database.Store, pub pubs
229237return xerrors .Errorf ("get provisioner job: %w" ,err )
230238}
231239
232- // Check if we should still unhang it.
233- if ! job .StartedAt .Valid {
234- // This shouldn't be possible to hit because the query only selects
235- // started and not completed jobs, and a job can't be "un-started".
236- return jobIneligibleError {
237- Err :xerrors .New ("job is not started" ),
238- }
239- }
240240if job .CompletedAt .Valid {
241241return jobIneligibleError {
242242Err :xerrors .Errorf ("job is completed (status %s)" ,job .JobStatus ),