mirror of
https://github.com/ipnet-mesh/meshcore-hub.git
synced 2026-07-26 11:33:12 +02:00
Louis King
cbeb6b38c2
Full-plan review surfaced 13 issues; all corrected in place and catalogued in review-findings.md (cross-referenced F1–F13). Four schema-level blockers resolved before the Phase 0 DDL freeze: - F1: instance-scoped composite uniques (nodes.public_key, event_hash, channels.name/key_hex) + settings PK (instance_id, key) — makes D21's "schema does not change" actually true - F2: only raw_receptions-sourced CAGGs remain; message/advert/node-count become worker-maintained dashboard rollup tables + dashboard-rollups job - F3: FORCE RLS + non-owner app role, per-request transaction for reads, instance_id in cache key, explicit instance_id predicate on CAGG reads - F5: D5 benchmark moved to Phase 0; Phase 1 decode shadow vs Phase 2 full parallel-stack Design-risk fixes: single INGEST stream + wildcard consumer (F8), stable two-arg advisory-lock key (F7), diff harness keyed on wire_hash (F4), FK-less hypertable node refs (F6), single-eval spam sweep (F10), observer upsert + telemetry dedup note + rowid chunk exclusion (F11). Notes for TS-translation gaps (F9), setup-wizard SSR (F12), and scope/risk (F13). Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>