Every PR stops paying for evals twice.
- Flakes are now measured, killed at the root, and fenced
- Every retried pass is now recorded in CI output where it cannot hide
- Free lane retries a failing file once and appends every flaky pass to a per-project ledger uploaded from CI on green runs
- Flake ranking command available via bun run eval:flake-rank
- Two-phase session timeouts
- Sync-spawn timeout tripwire for spawnSync, execSync, execFileSync, and Bun.spawnSync with comment-aware detection and 30-line window
- Behavioral kill-semantics tests
- CLI version stamping
- Fail-open reconcile gate
- Test infrastructure overhauled with derived gate census that covers new gates upon file landing instead of hand-enumerated rows
- Paid runner defaults changed from 4x4 to 8x2
- CI setup deduplicated into four composite actions
- Removed 499 timeout-less sync-spawn sites across 146 files that could cause hung child processes under blocking spawnSync
From gstack
Flakes are now measured, killed at the root, and fenced. The test infrastructure got its overhaul, wave 1. The legacy 17-row eval matrix that ran serialized AHEAD of the sliced lane on every PR is deleted: one paid lane, its gate census derived from the runner itself, so a new gate test is in the census the moment its file lands. No hand-enumerated rows to drift, and the drift already tried, a new matrix row landed on main mid-branch and the merge resolved to the derived census that covers it by construction. The flake war moved from anecdotes to instruments. Every retried pass is now recorded where it cannot hide (bun's own output shows a retry as a clean pass, we probed it), the free lane retries a failing file once, loudly, and appends every flaky pass to a per-project ledger uploaded from CI on green runs. bun run eval:flake-rank ranks the series. And the wedge class that hit main, a hung child under a blocking spawnSync that no in-process timeout can interrupt, is extinct: 499 timeout-less sync-spawn sites across 146 files (the branch tripwire's own count against main) swept to zero, with a ratcheted tripwire that failed its first real offender the day a timeout-less spawn arrived from a merge. Flake telemetry, end to end. Two-phase session timeouts. Green-by-skip census. Sync-spawn timeout tripwire (spawnSync / execSync / execFileSync / Bun.spawnSync, comment-aware, 30-line window, shrink-only exemption ratchet) plus a raw-SHA fixture ban. Behavioral kill-semantics tests. CLI version stamping. One paid lane. Paid runner defaults 4x4 to 8x2. CI setup deduplicated into four composite actions. Supply chain pinned. Routing journeys lost their answer key. Decided A/B experiments retired. A fail-open reconcile gate. A write-token trust boundary. Provider-runner orphans. Selection integrity. A cross-shard SKILL.md race.