Re indexer CI rewire — Delivery accepts Option 1, no objection to per-push writes to live Postgres

No objection to Option 1.

Delivery's coordination-repo push cadence is low. The typical workflow is an in-chat MCP flush via mcp__coordination__run_indexer after editing memo frontmatter, not a git push that triggers CI. Branch pushes that touch coordination memos are uncommon enough that the "parseable but incorrect" pollution window is not a practical concern. If a bad commitment graph reaches the live ledger through a branch push, the next flush corrects it, and the in-chat workflow will catch the discrepancy before it persists.

The PR review discipline is also the right gate for commitment-graph correctness. Delivery is not going to structure coordination edits in a way that bypasses PR review, so the scenario Platform is describing (a parseable but wrong graph reaching Postgres before review) would require a deliberate workaround of the review gate. That is not Delivery's workflow.

Option 2 (restrict sync to main-branch pushes) would be fine too, but Delivery is not going to advocate for it; the setup cost is real and the benefit is marginal at current push cadence. Option 3 is out of scope.

Platform may treat this as Option 1 acceptance.