Retry Google 499 CANCELLED as transient

[hiroshinishio]

Summary

• Added getattr(err, "code", None) == 499 to is_transient_error so Google's ClientError: 499 CANCELLED flows through the existing linear-backoff transient-retry path (2s/4s/6s, 3 attempts)
• Reproduced from the 2026-04-20T17:36:33 gitautoai/website incident (AGENT-3JX/3JY/3K0/3JZ): Google's backend cancelled the stream server-side during the same free-tier overload window that produced the 429 cluster 1 hour earlier, but without a Retry-After hint
• Added two tests using the real google_errors.ClientError shape: positive for 499 CANCELLED, negative for 400 INVALID_ARGUMENT (to prove we don't retry real client bugs)
• Minor cleanup: two partial-assertion fixups from the prior rate-limit PR and a sentry-cli doc update in CLAUDE.md

Social Media Post (GitAuto)

Google 499 CANCELLED errors now retry instead of bubbling to Sentry

• is_transient_error recognizes Gemini's server-side stream cancellations as retryable
• Linear backoff via the existing transient-retry budget (3 attempts, 2s/4s/6s), no new config
• Paired with the 429 retry shipped last week, free-tier overload windows now self-heal

Social Media Post (Wes)

Gemini's free-tier has two flavors of overload: 429 with a "retry in 60s" hint, and 499 CANCELLED with no hint at all. Same Sentry cluster, same repo, 1 hour apart. Four lines plus two tests to make 499 take the transient-retry path we already had. Overload windows self-heal now.