Geo FileDownload failures get retried only after finishing the backfill
Summary
If any failures happen during a backfill, the retries will only get picked up at the end of the backfill by the Geo::FileDownloadDispatchWorker
, as it only happens if it has remaining capacity (no other unsynced jobs).
What is the current bug behavior?
Described above, failures happening during a backfill are only retried after the backfill completes.
What is the expected correct behavior?
Not sure if we want to treat this differently, so this may end up as a documentation change - or potentially schedule retries earlier, intertwined with the backfill queue.
Initially encountered by Large Premium customer in ZD (internal-only)