Currently, job logs and results are uploaded to the public bucket when the job finishes (either success or failure). For long-running jobs, a user would reasonably expect to be able to check the job progress. Uploading the workspace at regular intervals (every few minutes?) should be sufficient for most situations.
Currently, job logs and results are uploaded to the public bucket when the job finishes (either success or failure). For long-running jobs, a user would reasonably expect to be able to check the job progress. Uploading the workspace at regular intervals (every few minutes?) should be sufficient for most situations.