📖About

Block-ownership audit (2026-06-09) found that of the 4 droids flagged as spanning pipeline blocks, 3 were over-reads (Merlin Replanner, Albus Reviewer, Bypass Shipper are each correctly single-block). The one genuine block-spanner is Elronds Handover Watchdog (el-handover-01): a 1579-line, ~9-gap scheduled watcher (gaps A,B,C,D,E,F,H,I,J + orphan/zombie/parent-remediation sweeps) that acts across the file->dispatch, review-initiative, plan/replan, judge, troubleshoot, and monitor boundaries. GOAL: redesign it into ~4 owner-scoped watchers so each watcher (and its gaps) sits in exactly one block under one owning agent. Proposed split: (1) Orion dispatch watcher = gaps A,E,H (un-dispatched/orphaned bypass enqueue, file->dispatch); (2) Elrond review/plan watcher = gaps B,C,F (reviewer-pending, planner-pending, bypass reviewer-close); (3) Albus troubleshoot watcher = gaps D,I (universal stuck detection -> Albus broadscan, Albus session cap); (4) Orion monitor watcher = gap J + orphan/zombie/remediation sweeps (execution cleanup). DESIGN-FIRST: do not implement until the split design, the timer/runner-script changes, and the registry droid rows are reviewed. This initiative captures the design; implementation follows via Orion bypass with twin-reviewer pass once the design is approved.

⚖️Decisions

• Superseded/absorbed by CAROL-INI-0924: the handover-watchdog split is realized as the Handover Gap Detector becoming a shared service, not 4 owned droids. (orion)
• [status-router] planned -> closed | event=operator_put | PUT /api/initiatives (operator)

✅Success criteria

• After the split, every resulting watcher droid maps to exactly one pipeline block; no droid spans blocks (must_have)
• Each new watcher is owned by the agent that owns its block (dispatch+monitor->Orion, review/plan->Elrond, troubleshoot->Albus) (must_have)
• All 9 existing gap behaviours (A,B,C,D,E,F,H,I,J + sweeps) are preserved with identical detection/action logic; verified by before/after behaviour parity (must_have)
• Timer, runner script, and registry droid rows are updated consistently and the fleet stays up after the swap (must_have)