08 - Scale and Performance Review

Scale Verdict

SCALE_NOT_SAFE

Measured Evidence

fn_dot_birth_qt001_plan_v2(system_issues), ~223,952 rows: 5.079 seconds. knowledge_documents ~6,631 rows: 0.220 seconds. fn_qt001_plan_all(), 74 collections: 4.942 seconds.

The planner executes three correlated birth-registry probes per source row plus collection count. At 100M rows this is not acceptable control-plane preflight.

Scale Breaks

Every batch rescans from beginning with no keyset/watermark. Multi-batch conflicts with one-use permits. No actual resume. Planner/writer row-by-row. Snapshot TRUNCATE/refill is not versioned/atomic and lacks PK/index/checksum/watermark/TTL. Counters use integer. No adaptive policy, load shedding, scheduler, monitor, stable source index contract, or concurrent insert reconciliation.

Required model: versioned set-based plans, stable keyset ranges, immutable watermark, per-range ledger, run lease, adaptive batch policy, and monitoring DOT.