[mberman84]

What changed

• keep draft and pending-review deploy requests waiting instead of falsely repair-blocking them
• create durable repair handoffs for stale exact-head intent and clean up abandoned thread records
• expose request-to-ready alerts and active pre-queue overlap groups
• add scheduled full-state reconciliation for coalesced GitHub event bursts
• batch and parallelize status and delivery-metrics reads while preserving exact ancestry checks
• update the DeployBot skill and reference documentation

Why

Live Astro deployment threads were repeatedly rebasing, silently waiting outside the queue, and carrying stale blocked/thread state. Status returned no useful alerts and metrics/status were too slow for operational use.

Validation

• ruff check src tests
• PYTHONPATH=src python3 -m unittest discover -s tests — 159 tests passed
• structured autoreview clean before the v0.2.9 rebase; post-rebase review had one self-retracted GraphQL-limit false positive verified against GitHub's 500,000-node limit
• live read-only Astro metrics: ~7.5 seconds, exact ancestry preserved
• live read-only Astro status: request-stage alerts and active-intent overlaps surfaced

Coordination

Thread 019eeaf2-d879-7173-8f90-75a2a641a984 confirmed it has not edited these areas and will rebase after this lands, then add only final merge-authorization revalidation, bounded repair-overlap protection, and the follow-default docs/test correction.