bench: attribute S2 grouping scheduler cost

[blove]

Summary

• add private scheduler-wait and one-transition grouping diagnostics
• partition the full interaction interval and support diagnostic-only budget perturbations
• commit the preregistered seven-block measurement milestone and excluded CDP trace

Result

At the 0.25 ms production budget, active row-model slice work is the dominant component at 77.1% of the full interval. Scheduler waits account for 14.1%; increasing the diagnostic budget to 2 ms reduces callback count by 77.8% but changes the full interval by only 1.4%. This does not justify a production budget change.

Verification

• pnpm run format
• pnpm run lint
• pnpm run typecheck
• pnpm test
• browser artifact contract for S2 smoke grouping diagnostics
• 28 accepted target-scale runs with all correctness and continuity invariants satisfied

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

1 Skipped Deployment

Project	Deployment	Actions	Updated (UTC)
pretable	Ignored		Aug 31, 2026 4:52pm

[github-actions]

Vercel preview ready

Preview: https://pretable-awzcos1op-cacheplane.vercel.app
Commit: ad5f694c2bea80d9ca55bf3a10e47120d5a30c45

_{Updated automatically by the deploy-preview job.}