mirror of
https://gh-proxy.org/https://github.com/klinsmaya/multica.git
synced 2026-11-08 17:36:45 +08:00
A provider rejection carrying the concurrent request limit witness is classified as provider_capacity_or_rate_limit ahead of the broad token/context and generic 403/auth rules, so members are no longer told to sign in again or reduce context for a transient concurrency rejection. NormalizeDaemonReason upgrades the same narrow witness from older daemons. Classification only: retryableReasons, retry delays and task creation are unchanged, and a regression pins provider_capacity_or_rate_limit as non-auto-retryable. The auto-retry proposal from #8180 is deliberately not included. Split from #8180 after review.