Troubleshoot dgrunner and runner jobs

First identify where the failing process runs: the Windows/Linux machine that hosts dgrunner, the Linux runner container, a GitHub-hosted runner, or a separate VPS deployment target. Do not run a permission repair on the VPS unless the failing path is actually on the VPS.

Classify the failure stage

1. Read the first runner error, not only later retry or disconnect messages. If the job failed before checkout, no checkout, build, or deployment workflow command ran. Later broker reconnect messages can be a consequence of a failed job rather than its cause.
2. Check the run's assigned runner, runner state, runs-on labels, and the exact failing path. A path like /home/runner/_work/_tool belongs to a Linux runner environment; determine whether it is inside the local Docker runner or a native Linux runner before changing ownership.
3. Collect read-only diagnostics first:

dgrunner doctor
dgrunner status
dgrunner ps
dgrunner logs --follow

For the default Docker runner on the local host: