release aidev PRERELEASE SIG 2/5

Gemini CLI nightly 20260808: capacity exhaustion now a terminal error

One user-facing change — capacity exhaustion is reclassified as a terminal error, so those failures stop being retried as transient. The rest of the merged work is the project's internal "caretaker" issue-triage infrastructure (Firestore schema, Cloud Run jobs, Pub/Sub, eval harness), which does not affect the CLI itself.

PUBLISHED2026-08-08
OBSERVED2026-08-11
AGE3d
SOURCES1
  • Capacity exhaustion reclassified as a terminal error (#28716)
  • Caretaker triage infrastructure: Firestore schema gains error and pr_number fields, triage Cloud Run job workflow, GCP deployment script, workable-spec event published to a ready-for-code Pub/Sub topic (#28467, #28468, #28529, #28588)
  • Caretaker evals: triage evaluation framework, judge runner, local golden-issue collection with Firestore sync, Cloud Run eval entrypoint (#28530, #28532, #28727)
  • Caretaker fixes: lock cleared on NEEDS_HUMAN transition; issue-comment handling and re-triage workflow (#28601, #28690)

COMMUNITY

No curated reactions recorded for this event. Facts and takes are kept in separate layers — community context is added by hand, never blended into the record above.