Resolves #13971

Related: #13755

Adds a backlog-aware REM catch-up path for saturated but non-overflowing Dream cycles. When a REM extraction run fills the configured batch limit, the scheduler can fire a bounded follow-up after the typed catch-up cooldown instead of waiting the full steady-state Dream cadence; non-saturated, overflowed, skipped, failed, or disabled-cooldown cycles keep the existing cadence.

Evidence: L2 (scheduler/service unit coverage for saturated REM cycles, cooldown gating, and completion-state metadata) -> L2 required (orchestrator scheduling contract). No residuals.

Deltas from ticket

• Uses AiConfig.orchestrator.intervals.remBacklogCatchupCooldownMs as a typed leaf and fails loud when registry wiring omits it.
• Persists Dream completion metadata in TaskStateService so the scheduler can distinguish saturated REM work from ordinary successful Dream runs.
• Keeps the existing hourly Dream cadence for non-saturated cycles and for cycles that exceeded the runtime threshold.

Test Evidence

• git diff --check
• git diff --cached --check
• npm run test-unit -- test/playwright/unit/ai/daemons/orchestrator/scheduling/dream.spec.mjs test/playwright/unit/ai/daemons/orchestrator/scheduling/registry.spec.mjs test/playwright/unit/ai/daemons/orchestrator/scheduling/pipeline.spec.mjs test/playwright/unit/ai/daemons/orchestrator/services/TaskStateService.spec.mjs test/playwright/unit/ai/daemons/orchestrator/services/DreamService.executeRemCycle.spec.mjs -> 62 passed.

Post-Merge Validation

• Restart the orchestrator and confirm a saturated REM cycle schedules rem-backlog-catchup after NEO_ORCHESTRATOR_REM_BACKLOG_CATCHUP_COOLDOWN_MS instead of waiting dreamMs.

Commits

• 2dd5e7c7e5 — fix(ai): add REM backlog catch-up scheduling (#13971)

Authored by Euclid (GPT-5, Codex Desktop). Session 019ef378-527d-7393-bc74-ec3a1d3f2ddf.