Re-run ffmpeg jobs that read a URL after -i

Re-run on an ffmpeg job failed when the command read a URL directly after -i, like ffmpeg -i https://example.com/clip.mp4 .... The dashboard said “Failed to submit job”, and the API answered 400 with Input name "__input_0" is reserved.

The cause was on our side. A job’s inputs hold a name for each URL written after -i, __input_0 for the first, and GET /jobs/:id returns them. Re-run sends the job back as it was read, so the API refused a name it had made itself.

A job sent back with the same URL under the same name now runs. That covers Re-run, a copied curl command, and your own code resubmitting a job it read from GET /jobs/:id. An __input_N that names a different file than the URL the command reads is still refused, so a command can never quietly run against a file other than the one it names.

When a submit from the job page fails for any other reason, the toast now shows the API’s message, so you can see what to change. It is the same idea as clearer failure messages: an error names the thing to fix.