Repository navigation
Issue-sweep-rotating endgame: convert to a Make-role triage feeder, or retire #2138
Description
issue-sweep-rotating (hourly, one repo per run, up to 5 items) works open issues directly, outside the queue system. It predates the queue pilot and now overlaps the opencode/openchamber queue lanes — two systems working the same issues, kept apart only by labels.
Decide and execute one endgame: (a) convert the sweep into the Make role's triage feeder — it stops working issues itself and instead enqueues them into the pilot queues with triage packets; or (b) retire it once the session conductor covers those repos. Either way, end the parallel non-queue execution path. Record the decision and the deciding evidence (per-item costs from the token check-ins) on this issue.
Activity
developing-today commentedon Oct 29, 2026
CODE-1036
dezren39 commentedon Oct 29, 2026
TOP PRIORITY (Drew, 2026-10-29 01:23 CDT): Batch 3 fleet consolidation — pick this up in the next runs, ahead of routine work. Being enqueued to repo-health-code-pilot at top priority now.
dezren39 commentedon Oct 29, 2026
🤖 Agent note · picked up
dezren39 commentedon Oct 29, 2026
🤖 Agent note · done
Decision: (a) CONVERT to the Make-role triage feeder — do not retire yet
Premise check (live, today): the parallel path is not theoretical. This morning's
sweep.log(2026-10-29) showsissue-sweep-rotatingdirectly fixing opencode issues outside the queue — PR #836 for opencode#829 and PR #838 for opencode#831 at 06:36Z — and, decisively, queue task 3137 had to be cancelled as self-assigned because the sweep took its item: the two systems collided on live work today, kept apart only by after-the-fact label surgery. The opencode/openchamber/nix queue lanes are all enabled (live scheduler), so the overlap the issue describes is the fleet's current state, not a risk.Cost evidence: the weekly token review's baseline table puts
issue-sweep-rotatingat 0.660B input for the baseline week — the 4th-largest cron job in the fleet — spent substantially on direct execution that the queue lanes now perform under claim discipline. (Source: weekly-token-usage-review job record, baseline week of the 2026-10-22 scheduled-tasks review; check-in #5's per-job table, referenced by this issue, is the fresher cut the implementation's first cost report should use.)Why not retire (b) now: the retire condition — "once the session conductor covers those repos" — is not met and is partly misconceived: the conductor (#2153, designed, not yet built) dispatches queued work; it does not triage raw backlogs. The sweep's one unique contribution is exactly that triage judgment over the opencode/openchamber/nix backlogs. Retiring today reopens the triage gap for three repos and throws away the judgment while keeping none of the discipline benefit — the queue lanes would simply receive less, later. Conversion keeps the judgment, ends the unclaimed execution, and makes the sweep's output measurable as packets enqueued per run.
The conversion in one paragraph: hourly rotation and up-to-5 volume unchanged; the sweep triages and enqueues into each repo's pilot queue (with dedupe against live tasks and
work/*labels before every enqueue, andqueue/queuedset in the same pass); it never fixes, never opens PRs, never self-assigns; clerical closes with evidence (already-fixed/duplicate) remain triage and stay allowed. Retire-revisit trigger: two consecutive weekly reviews showing feeder packets substantially duplicating producer coverage, or conductor + producers demonstrably keeping all three backlogs triaged without it → retirement returns here as a fresh decision, not by drift.Routing
Implementation filed as #2156 (feeder-only body rewrite, dedupe rule, three evidenced feeder runs, role-map update, first per-item cost report) and enqueued as queue task 3180, flagged
live-fleet-change. Role note: confirms the sweep's Make placement in the #2131 map — as a feeder, which is what that map already calls it.developing-today commentedon Nov 3, 2026
agent-note:v1
state: reviewed (coverage loop-back)
Coverage pass 2026-11-03, loop-back review: this issue was re-reviewed (drift / overlap / already-handled / stale / blockers / ordering) so every open issue in the four repos is reviewed and in the system. Verdict: still live; in the system as task 7302 (triage, queued, priority 50) in repo-health-code-pilot. Labels audited and synced. If it is already handled at current main, its triage closes it with evidence; if oversized, the triage files the split; blockers belong on this issue.
Found via: coverage loop-back 2026-11-03
dezren39 commentedon Nov 3, 2026
🤖 Agent note · picked up
dezren39 commentedon Nov 3, 2026
🤖 Agent note · done (triage)
Triage verdict (2026-11-03, task 7302): decision executed; exactly one item remains, already tracked
Premise re-verified: this issue asked for a decision (convert vs retire) plus execution. Both halves are done:
issue-sweep-rotatinginto the Make-role triage feeder, with evidence (live collision with queue task 3137 that day; sweep at 0.660B input in the baseline week, 4th-largest cron) and an explicit retire-revisit trigger.sweep.loglines at 2026-11-03T21:47Z–23:38Z show the sweep still operating enqueue-only — enqueued nix triage tasks 8909, 8910, 8912 withqueue/queuedlabels and dedupe-skips recorded — no direct fixes, no PRs, no self-assignment. The parallel non-queue execution path this issue exists to end is ended.What remains (the only blocker, already owned): the decision requires the deciding evidence to include per-item costs, and the first post-conversion weekly-review cost-per-item report has not yet landed on this issue. That is the recorded park condition of task 3808 (parked, on #2156): when the report lands there, task 3808 completes and this issue can close citing both evidence halves. Nothing else is owed: no split, no new packet — filing a duplicate follow-up here would recreate the drift this loop-back exists to prevent.
Ordering: #2138 stays OPEN until the cost report lands; the retire-revisit trigger (two consecutive weekly reviews showing feeder packets substantially duplicating producer coverage, or conductor + producers keeping all three backlogs triaged without it) is unchanged and unevaluated until the first report exists.