# Runs, retries, and cancellation

Follow a build from upload to review and recover when capture fails.

## Find a run
The project run list shows commits, branches, pull requests, status, and snapshot usage. Open the result associated with the commit you intend to review. A later upload may supersede an older run, and native projects can group multiple platforms for one commit.

## Understand progress
Queued means the upload is accepted and waiting. Running means capture is in progress. Completion means processing finished; it can still have changes to review or story errors. Check the visual result and policy conclusion, not just whether the process completed.

A run that failed before producing trustworthy comparisons is not evidence that the component is unchanged.

## Re-run failed stories
Where the run offers **Re-run errored stories**, confirm the scope before starting it. It creates a new run that captures those errored stories again and carries over the other results. Re-running an unchanged uploaded build can help with a transient failure; it cannot include a source fix that was never built and uploaded.

Other re-run actions explain their scope in the confirmation. A retained build and suitable run state are required. A merged or superseded pull-request run may not offer the same actions as the current run.

## Automatic retries
Framecheck may retry after a problem on its side. The run reports that it is retrying, and you are not charged extra for that retry. Do not assume every user-triggered new capture is the same as an automatic recovery; read its confirmation and usage summary.

## Cancel
If offered for an active run, use Cancel to stop work you no longer need. Cancellation is not a pass. Work already performed can count as usage; queue time and work never performed do not. Cancelling a job in your CI after its upload has queued a run is not necessarily the same as cancelling that run in Framecheck.

## Decide between re-run and new upload
Use a re-run for the same build when the app offers it and the cause was transient. If you changed a story, dependency, build output, or capture entry, build the new commit and upload it. Keep the commit metadata accurate so reviewers can tell which code produced the result.
