GFS Cron Schedule — R121 Tightened Cadence + R583 reconciliation

31 cron triggers across 11 cadence bands · audited 2026-05-25 against wrangler.jsonc (R583)
⚡ Every 2 min · */2 * * * * · R121 TIGHTENED (was */5) 720 cron invocations/day · NS governance ~25% per cycle · Hot table writes land in D1 ≤ 2 min handleSyncTier(env, 'hot') transactions / invoice_lines / so_lines handleSystemNoteCDC(env) NetSuite SystemNote change-data-capture 🌡 Every 5 min · */5 * * * * · R121 TIGHTENED (was */15) 288 cron invocations/day · NS governance ~15% per cycle · Master records refresh 3x more often handleSyncTier(env, 'warm') customers · vendors · items master refresh _pricingMasterRefresh(env) incremental pricing_master sync (R72.1) 🪞 Every 15 min · */15 * * * * · R121 TIGHTENED (was */30) 96 cron invocations/day · Low NS impact · Verification + session summaries land sooner handleMirrorBackVerify(env) ADR-019 NS push verify · drift detection handleSessionSummarizerCron(env) R107 · idle 30min threshold · Haiku 4.5 ⏰ Hourly · 0 * * * * 24 cron invocations/day · Cold-tier records (rarely change) handleSyncTier(env, 'cold') ⏱ Every 2 hours (offset +5min) · 5 */2 * * * 12 cron invocations/day · Healthcheck framework handleHealthcheckDispatch(env) ⏳ Every 4 hours · 0 */4 * * * 6 cron invocations/day · Review queue SLA escalation handleReviewQueueSlaCron(env) 📅 Daily · 5 triggers 5 cron invocations/day · Nightly caches · daily reports · red-team probes · line backfill · pricing discrepancy 0 2 * * * — line backfill handleLineTableBackfill (R95) 0 3 * * * — cache warm handleNightlyCacheWarm (R45) 0 6 * * * — pricing drift handlePricingDiscrepancyDetect (R72) 0 7 * * * — red-team runRedTeamProbes (R75-G · ~$0.21/run) 0 19 * * * — daily report handleDailyReport 🗓 Weekly · 3 triggers 3 cron invocations/week · Backups + analytics export + llm-wiki rebuild 0 3 * * SUN — llm-wiki handleLlmWikiRebuild (R84) 0 4 * * SUN — backup + export D1 backup + analytics CSV export (R88) 0 8 * * SUN — learning promote handleWeeklyLearningPromote (R47) 📆 Monthly (1st of month) · 2 triggers 2 cron invocations/month · Archival + source backup 0 3 1 * * — handleArchival 0 5 1 * * — handleMonthlySourceBackup 🆕 Recent additions (R122 → R566) · 15 triggers (R583 reconciliation) Karpathy eval · drift triage · learning maint · anomaly hunter · digest · training · grader · jarvis · health snap · cost basis · vendor items · sync recon · verify scheduler · corpus rebuild · customer health 7 * * * * — eval cron handleEvalReplayCycle (R122) · 50 cases/h 22,37,52 * * * * — drift triage handleDriftAutoTriage (R174/175) 8,38 * * * * — verify scheduler R563 · workflow_verify_results drain 17,47 * * * * — jarvis watchdog R347 · stale crons + HITL aging 12 */4 * * * — training loop R259 · 5 questions sampled/4h 30 3 * * * — learning maint R175 · re-embed decision_corpus + KV warm 30 6 * * * — morning digest R201 · 2:30am ET digest email 45 5 * * * — health snapshot R531 · system_health_snapshots row 0 6 * * MON — anomaly hunter R181 · 6 variance queries → weekly_anomalies 0 1 * * SUN — cost basis R534 · stages bulk_cost_basis proposals 30 1 * * SUN — vendor items R536 · backfill_vendor_items_netsuite_id 30 2 * * SUN — cost rollup R533 · assembly_cost_rollup recompute 15 4 * * SUN — corpus rebuild R563 · knowledge_chunks re-embed (placeholder) 30 4 * * SUN — tool grader R285 · weekly partial/fail re-grade 45 4 * * SUN — sync recon Phase 48 · NS vs D1 drift on 6 master tables

R121 — what changed

  • • Hot tier */5 → */2 — transactions land 2.5x faster
  • • Warm tier */15 → */5 — masters refresh 3x more often
  • • Mirror-back */30 → */15 — push verification 2x faster
  • • Old `else` fallback dropped — unknown crons now log a warning

Total cost

  • • ~1,500 cron invocations/day total
  • • Well under CF free tier (100K/day)
  • • NS governance throttled by CostCapDO
  • • Red-team probe ~$0.21/run = ~$6/month

Per-cycle hot signals

  • */2 · hot tier — transactions / invoice_lines / so_lines
  • */5 · warm tier — customers / vendors / items master
  • */15 · mirror-back + R107 session summaries
  • 0 * · cold tier — rarely-changing records

Daily orbit

  • • 02:00 UTC · line-table backfill (R95)
  • • 03:00 UTC · nightly query-cache warm (R45)
  • • 06:00 UTC · pricing discrepancy detect (R72)
  • • 07:00 UTC · red-team probes (R75-G)
  • • 19:00 UTC · daily report

Weekly + monthly

  • • SUN 03:00 · llm-wiki rebuild (R84)
  • • SUN 04:00 · D1 backup + analytics CSV export (R88)
  • • SUN 08:00 · learning promote (R47)
  • • 1st 03:00 · monthly archival
  • • 1st 05:00 · monthly source backup

Governance budget

  • • CostCapDO enforces per-cycle NS budget
  • • Hot tier ~25% of cycle budget
  • • Warm tier ~15% of cycle budget
  • • Daily probes/reports under 5% combined
  • • Plenty of headroom for R122 eval cron