chore(deps): update experiments digest to 974c590

[renovate-fullsend]

This PR contains the following updates:

Package	Update	Change
experiments (changelog)	digest	c4c3e79 → 974c590

---

Configuration

📅 Schedule: (UTC)

• Branch creation
  • At any time (no schedule defined)
• Automerge
  • At any time (no schedule defined)

🚦 Automerge: Enabled.

♻ Rebasing: Whenever PR is behind base branch, or you tick the rebase/retry checkbox.

🔕 Ignore: Close this PR and you won't be reminded about this update again.

---

• If you want to rebase/retry this PR, check this box

---

This PR has been generated by Mend Renovate.

[fullsend-ai-review]

🤖 Finished Review · ✅ Success · Started 3:23 PM UTC · Completed 3:28 PM UTC

Commit: e9ca75d · View workflow run →

[github-actions]

Site preview

Preview: https://bc2ef1ba-site.fullsend-ai.workers.dev

Commit: 0addf1b5fe92398719e2403ed0be4c06d886cfb1

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!

[fullsend-ai-review]

Looks good to me

Previous run

Looks good to me

Previous run (2)

Looks good to me

Previous run (3)

Looks good to me

Previous run (4)

Looks good to me

Previous run (5)

Looks good to me

Previous run (6)

Looks good to me

Previous run (7)

Looks good to me

[fullsend-ai-review]

🤖 Finished Review · ✅ Success · Started 3:29 AM UTC · Completed 3:34 AM UTC

Commit: ac4422e · View workflow run →

[fullsend-ai-review]

🤖 Finished Review · ✅ Success · Started 3:21 PM UTC · Completed 3:26 PM UTC

Commit: 53bbaf6 · View workflow run →

[fullsend-ai-review]

🤖 Finished Review · ✅ Success · Started 3:33 AM UTC · Completed 3:41 AM UTC

Commit: c6f5725 · View workflow run →

[fullsend-ai-review]

🤖 Finished Review · ✅ Success · Started 3:21 PM UTC · Completed 3:29 PM UTC

Commit: 8b6f136 · View workflow run →

[fullsend-ai-review]

🤖 Finished Review · ✅ Success · Started 3:35 AM UTC · Completed 3:43 AM UTC

Commit: fc1b1da · View workflow run →

Runtime: claude · Model: opus → claude-opus-4-6 · Effort: high · Cost: $1.50

[fullsend-ai-review]

🤖 Finished Review · ✅ Success · Started 3:27 PM UTC · Completed 3:33 PM UTC

Commit: f47b277 · View workflow run →

Runtime: claude · Model: opus → claude-opus-4-6 · Effort: high · Cost: $1.65

[fullsend-ai-review]

🤖 Finished Review · ✅ Success · Started 3:31 AM UTC · Completed 3:39 AM UTC

Commit: 0addf1b · View workflow run →

Runtime: claude · Model: opus → claude-opus-4-6 · Effort: high · Cost: $1.32

[fullsend-ai-retro]

🤖 Finished Retro · ✅ Success · Started 9:18 AM UTC · Completed 9:27 AM UTC

Commit: 0addf1b · View workflow run →

Runtime: claude · Model: opus → claude-opus-4-6 · Effort: high · Cost: $2.90

[fullsend-ai-retro]

Retro: PR #6453 — Renovate submodule bump reviewed 8 times

PR #6453 was a trivial 1-line Renovate bot PR bumping the experiments git submodule pointer. The review agent (fullsend-ai/agents@v0.37.0, claude-opus-4-6, high effort) ran 8 times over 4 days (Aug 21–25), each triggered by a Renovate rebase. Every run dispatched the full review pipeline (correctness + style-conventions sub-agents), and every run approved with "Looks good to me" — zero findings across all 8 runs. Estimated total review cost: ~$10–12 for a change that never varied.

A human reviewer approved on Aug 25 at 09:15 UTC and the PR merged one minute later.

Evidence for existing issues (no new proposals)

All findings are covered by existing open issues:

• Review agent should avoid full re-reviews when Renovate rebases without content changes #4596 (Review agent should avoid full re-reviews when Renovate rebases without content changes): This PR is a textbook instance — 8 identical LGTM approvals on an unchanged diff, ~$10–12 in wasted compute. The review agent checked prior reviews each run (found 7 prior LGTMs) but had no fast-path to skip re-analysis when the diff was content-identical.
• Review sub-agents should fall back to an available model when the primary model is unavailable #1771 / Review orchestrator passes roster-table model alias instead of frontmatter model to sub-agents agents#256 (Sub-agent model availability fallback): The style-conventions sub-agent failed because claude-sonnet-4-5@20250929 was not available on the Vertex deployment. The orchestrator recovered by retrying with the default model, but this added latency and token cost. The underlying model-routing issue persists.
• Consolidate 22+ overlapping "skip retro/agent dispatch" issues into a single design issue #5817 (Consolidate 22+ overlapping skip-dispatch issues): This retro itself is part of the pattern — a full retro analysis on a trivial bot PR with predictable findings. While the PR had agent involvement (8 review runs), all were trivial approvals, making the retro low-value.

Workflow quality

• Review quality: Appropriate — the correctness sub-agent verified the submodule's usage in test infrastructure and Go test files. No false positives or missed issues.
• Rework rate: Zero — no changes requested, no iterations needed.
• Token cost: The primary concern. ~$10–12 spent on 8 identical reviews of a 1-line change. Implementing Review agent should avoid full re-reviews when Renovate rebases without content changes #4596 (diff-hash fast-path) would reduce this to ~$1.50 (one full review + 7 fast re-approvals).
• Time to resolution: 4 days from PR open to merge, gated on human approval, not agent review.
• Autonomy readiness: The human reviewer's approval added no analytical value beyond the agent's 8 prior approvals — this class of change (bot-authored submodule bumps) is a strong candidate for full agent autonomy. See Track autonomy readiness: review agent on PR #4102 fully covered human review findings #4852 and Track autonomy readiness: review agent fully covered human review on code-agent-authored SHA bump in konflux-ci PR #8133 #4744 for related tracking.