Apps that serve field workers in Thailand often grow an offline queue early. The first version retries forever. The second version still retries forever, only with a nicer spinner.
Questions we ask in audits
- Is there a maximum retry count and a jittered backoff?
- Do retries respect OS background limits on both platforms?
- Can a poisoned payload block the entire queue?
- What does the user see when the queue is stuck for a day?
Why it becomes a release issue
A retry storm rarely fails a unit test. It fails a fleet. During one logistics review we tied elevated API 429s to a single Java uploader with no ceiling. The fix was small; the detection required reading production-adjacent code paths, not the happy-path demo script.
If offline is core to your product, mention it in the audit intake. We will allocate review time there instead of polishing screens that never leave Wi-Fi.