Repository navigation
[Bug]: Infinite retry loop on consistent provider failure #29143
Description
Description
The fallback system can spin in an infinite retry loop when a provider consistently fails. We should cap retry attempts (e.g. at 5) so the fallback system can proceed to the next configured model instead of spinning indefinitely.
Steps to Reproduce
- Configure a provider that consistently fails (e.g., due to network error or repeated 500 responses).
- Trigger the fallback system to use this provider.
- Observe the session logs showing an infinite retry loop without proceeding to the next configured model.
Expected Behavior
The retry loop should be capped at a reasonable limit (e.g., 5 attempts). Once the limit is reached, it should return an exhaustion status so the system can proceed to the next fallback model.
Environment
- OS: Any
- OpenCode version: dev
- Provider/model: Any consistently failing provider
Metadata
Metadata
Assignees
Labels
No labels
Activity
github-actions commentedon May 24, 2026
This issue might be a duplicate of existing issues. Please check:
github-actions commentedon Jul 24, 2026
To stay organized issues are automatically closed after 60 days of no activity. If the issue is still relevant please open a new one.
niStee commentedon Aug 13, 2026
Reopened for reassessment: the linked contributor PR was closed by automated stale-PR cleanup rather than merged. The issue should be re-triaged against current
dev; a refreshed PR should follow the current contribution template and include current verification.niStee commentedon Aug 13, 2026
Correction: GitHub rejected my reopen request because this account lacks repository permission to reopen issues. The issue remains closed. The recommendation is a fresh issue or maintainer-assisted reopen only if the behavior is still reproducible on current
dev; the old PR should not be reopened unchanged.