[KAJdev]

A job that was already accepted could fail with "not deployed" when its status lookup kept returning 404.

job = transcribe.spawn(url)
job.status()

Before, after seven 404s from the status route (about 32 s of retries):

runpod.apps.errors.EndpointNotFound: 'transcribe' in app 'demo' is not deployed. run `rp flash deploy` first, or use .local() to run it here.

The job was admitted, so the endpoint exists. A 404 on a status lookup only says the job wasn't found.

Now:

runpod.apps.errors.JobStatusError: could not retrieve status for job 'abc-123': HTTP 404 (GET https://api.runpod.ai/v2/flash/status/abc-123)

JobStatusError is an AppError exported from runpod.apps, with job_id, status, method and url attributes. The URL has any credentials, query string and fragment removed, and the message never includes request headers or the response body, so no API key or task token ends up in it.

It applies wherever the SDK reads a job's status: job.status(), job.result(), and the polling behind .remote(). Other HTTP errors on a status lookup (401, 503 and so on) now raise JobStatusError too, instead of a bare aiohttp.ClientResponseError.

Retries are unchanged. A status 404 is still retried up to seven times with the same backoff, and a failed submission is still never resent. A 404 on the submission itself still raises EndpointNotFound, and so do cancel, retry and stream.

Refs CON-2066.