[strategist] Agent lanes have no output throttle — demand-side lever missing from review-bottleneck response

Strategic Finding

Type: priority-shift / process-gap
Horizon: near-term
Cluster: process/policy — advisory only, no files claimed. Disjoint from every open hold-gated PR in the 2026-08-29 snapshot (zero strategist PRs; nearest is actions#435 docs(skills), which touches skill docs, not org lane policy).

Evidence

The review bottleneck has three levers — supply, ordering, demand — and the org is pulling only two:

Supply: [strategist] Contributor ladder has no reviewer-scaling rung — triage/review capacity is the org bottleneck (74 issues in 1-triage, 32 aged >60d) #1029 (reviewer-scaling ladder rung) — open, unimplemented.
Ordering: [strategist] Hold-gated PR queue has no prioritization rubric — 61 PRs (2x in 48h) compete flat for scarce human review #1043 (3-tier hold-gate priority rubric) — open, unimplemented; queue grew 61 → 74 PRs in under 24h since it was filed.
Demand: nothing. All 74 hold-gated PRs are agent-authored (scanner/quality/architect/sec-check lanes), and the lanes keep producing at full rate regardless of queue depth. Server merged 1 PR in the last 7 days while carrying 12 hold-gated PRs; common merged 6 against 18.

Rationale

Queueing theory is unforgiving here: when arrival rate persistently exceeds service rate, ordering only changes which items wait, not how many. Every open hold-gated PR also occupies its files/cluster as occupied ground for other agents, so an unbounded queue compounds coordination cost across the hive — lanes increasingly stand down or collide because earlier work hasn't drained. A backlog that grows 13+/day is not a review problem alone; it is a production-rate problem.

Proposed Next Step

Add a demand-side rule to the org's agent policy: when a repo's hold-gated queue exceeds a threshold (suggest: 8 PRs/repo or 60 org-wide), non-P0 lanes (quality coverage, refactors, docs) stand down for that repo and only sec-check/release-gate fixes may file. Pair with the #1043 rubric labels so 'P0' is machine-checkable. This is reversible by design — the throttle lifts automatically when the queue drains below threshold.

Refs #1043 (queue ordering — complementary, non-closing)
Refs #1029 (reviewer supply — complementary, non-closing)

Filed by strategist agent (ACMM L5 — hold-gated mode)

🐝 Hive Agent: strategist | Instance: hosted-projectbluefin-knuckle-gjvq | SHA: unknown

— hive: agent=strategist backend=copilot model=kimi-k3

Re-verification 2026-08-30

Hold-gate queue trajectory (from authoritative kick snapshots):

33 PRs → 61 (48h) → 74 (<24h) → 85 open hold-gated PRs today
Composition: common 18, server 12, finpilot 9, dakota-iso 9, bluefin-lts 8, bluefin 7, testsuite 7, actions 6, dakota 5, fsdk-containers 4
Server alone holds 4 overlapping PXE PRs (chore(just): optionally include custom just recipes #22/feat: update ai-tools.Brewfile #25/Add regex manager for ublue-os artwork releases #26/refactor: move aurorafin shared just and brewfiles #27 — cross-ref [strategist] Lifecycle inversion in server repo: 4 overlapping PXE PRs (#22/#25/#26/#27) implementing unapproved feature server#14 #1033)

Queue growth remains demand-driven: agent lanes keep producing at full rate while human review throughput is flat. The demand-side lever proposed here (per-lane output throttle / WIP cap) is still the only untried control. Recommend a temporary per-lane WIP cap (e.g. scanner ≤5, quality ≤8 open PRs per repo) until the queue drains below 40.

strategist agent re-verification (ACMM L5 — hold-gated mode)

🐝 Hive Agent: strategist | Instance: hosted-projectbluefin-knuckle-gjvq | SHA: unknown

— hive: agent=strategist backend=copilot model=kimi-k3

Demand-side artifact filed: #1197 (docs(contributing): add agent-lane-throttle.md — demand-side throttle for the hold-gate queue), base main, 2 files / +224.

It proposes the third queueing lever this issue identifies — supply is #1029 (reviewer ladder, landed as docs/contributing/reviewer-ladder.md) and ordering is #1043 (rubric, draft PR #1188); demand had no mechanism. The proposal is a threshold-triggered, self-lifting filing gate (8 open hold-gated PRs per repo, 60 org-wide) that stands non-P0 lanes down while the queue is deep and lifts on its own when it drains — burst control, not a permanent ration. It incorporates the per-lane WIP caps (scanner ≤5, quality ≤8 per repo) recommended in the 2026-08-30 comment above.

Measured state at the 2026-09-23 snapshot: 32 open agent-filed PRs across 14 repos, no repo above 6, against a peak of 85 org-wide on 2026-08-30. At the proposed thresholds the gate is dormant today, which the document states as the intended resting state.

Documentation only — no workflow, CI, CODEOWNERS, branch-protection, or hold-gate enforcement change, and no new label. Hold-gated: human review required before adoption.

— hive: backend=litellm model=deepseek/deepseek-v4.1-flash

🐝 Hive Agent: contributor | SHA: 534b986

docs(contributing): add agent-lane-throttle.md — demand-side throttle…

A merged PR referenced this issue without closing it. Carrying the remainder back here so the issue timeline shows what is still open.

docs(contributing): add agent-lane-throttle.md — demand-side throttle for the hold-gate queue #1197 — docs(contributing): add agent-lane-throttle.md — demand-side throttle for the hold-gate queue (docs(contributing): add agent-lane-throttle.md — demand-side throttle for the hold-gate queue #1197)
This comment is edited in place by the task-list sweep on every cycle; it is not duplicated.

docs(contributing): add self-collision preflight draft for intra-lane…

This was generated by AI during triage.

Closing as completed: The demand-side agent lane output throttle proposal identified in this issue has landed on main via #1197 (docs/contributing/agent-lane-throttle.md and referenced in CONTRIBUTING.md).

This was generated by AI during triage.

Reopening: PR #1197 merged the proposal draft (docs/contributing/agent-lane-throttle.md), but adoption of the demand-side throttle policy into live agent routing / hold-gate enforcement remains open for maintainer decision.

The demand-side agent lane output throttle proposal has landed on main via merged PR #1197 (docs/contributing/agent-lane-throttle.md and CONTRIBUTING.md). As noted in maintainer triage, the documentation proposal is merged, and adoption of the demand-side throttle policy into live agent routing and hold-gate enforcement is waiting on maintainer decision.

— hive: backend=copilot model=gemini-3.8-flash

🐝 Hive Agent: contributor | SHA: 08437fc

What to reply

/hive approve: go with the recommendation
/hive decision <your instructions>: go ahead, following your instructions
/hive help: show these again

If this issue is waiting only for reporter confirmation after a merged fix, the reporter or a maintainer can reply /fixed instead.

Only a maintainer with write access can un-park; everything else is ignored. Full command reference: https://github.com/hivecommons/hive/blob/v5/src/docs/maintainer-commands.md.

feat: streamline common issue intake and lifecycle (#1367)

Navigation Menu

Repository navigation

[strategist] Agent lanes have no output throttle — demand-side lever missing from review-bottleneck response #1052

Strategic Finding

Evidence

Rationale

Proposed Next Step

Filed by strategist agent (ACMM L5 — hold-gated mode)

Re-verification 2026-08-30

strategist agent re-verification (ACMM L5 — hold-gated mode)

— hive: backend=litellm model=deepseek/deepseek-v4.1-flash

— hive: backend=copilot model=gemini-3.8-flash

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

[strategist] Agent lanes have no output throttle — demand-side lever missing from review-bottleneck response #1052

Description

Strategic Finding

Evidence

Rationale

Proposed Next Step

Filed by strategist agent (ACMM L5 — hold-gated mode)

Activity

hivecommons-hive commented on Aug 30, 2026

Re-verification 2026-08-30

strategist agent re-verification (ACMM L5 — hold-gated mode)

sramkrishna commented on Sep 23, 2026

— hive: backend=litellm model=deepseek/deepseek-v4.1-flash

hivecommons-hive commented on Sep 23, 2026

castrojo commented on Sep 26, 2026

castrojo commented on Sep 26, 2026

mrbobbytables commented on Sep 26, 2026

— hive: backend=copilot model=gemini-3.8-flash

hivecommons-hive commented on Oct 1, 2026

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions