The stream for unsettled, the ledger for settled.

The same coverage predicate that drives replay retires the outbox: an intent is retired when its name carries an applied version at least as new — coverage, not acknowledgment. After the loss drill closed at 50 of 50, all 50 intents are deletable, the applied memory keeps all 50 names, replay finds nothing, and a reopen still remembers. The committed gate:

derive (compaction): an intent is retired when its name carries an applied version at least as new -- coverage, not acknowledgment -- so after H4 all 50 intents are deletable, the applied memory keeps all 50 names, replay finds nothing, and a reopen still remembers
H5 compaction ok -- all 50 intents retired by coverage in one pass, the applied memory kept its 50 names, replay over the compacted journal found nothing to do, and a fresh open of the same file still answers the last word -- the outbox empties, the memory does not
PASS 6/6

compact(j) is the retention verb, idempotent and safe to run on a schedule; the outbox stays small, the last word per name persists. Intents are small, coverage is one SQL pass, and an outbox kept under a few thousand rows replays in milliseconds.

The journal's durability is not free, and the record prices it exactly. On prepared-once statements — the single writer's privilege, with bind resetting the statement — the writer's pair is two WAL commits and one cached rowid read at synchronous=NORMAL. The committed gate:

derive (price): on prepared-once statements -- the single writer's privilege, with bind resetting the statement -- the writer's pair is two WAL commits and one cached rowid read at synchronous=NORMAL, so expect between 20 and 250 us on this disk; the remembered lane's end-to-end median should land between 200 us and 2 ms against the bare lane's committed 148 us: dearer, bounded, and the chapter's reason the journal is declared per group rather than assumed
H6 price ok -- the memory's price on this disk: 143 us per record-and-mark pair at the writer's edge, and a remembered lane end-to-end median of 524 us against the bare lane's committed 148 us -- 3.5 times the latency buys an outbox, a last word per name, and a replay that survives the bus
PASS 6/6

The priced pair is 524 µs (the remembered lane's end-to-end median) against the bare lane's committed 148 µs — the two numbers travel together — and 143 µs is the record-and-mark pair at the writer's edge. At synchronous=NORMAL the WAL issues no per-commit sync — the checkpoint is the sole I/O barrier — which is the journal's speed and its stated power-loss trade. The numbers carry their header: one scheduler, this container's disk, a 29-byte payload; the 3.5 multiple travels better than the 143.

The settled side is the design's Trading.Ledger (status PROPOSED). Double-entry in Postgres: branded id columns under the SQL canon with the domain's reject table as the floor, every posting one Ecto.Multi — a single all-or-nothing transaction, so a fill either posts both entries or neither. The directional decision is named in the spec and revisitable only by editing the spec: the stream (the journal, then milestone-B stream lanes) is the source of truth for unsettled state; the Postgres ledger is the regulated record of settled positions.

Exchange.Projection (also PROPOSED) is the idempotent log consumer that folds the log into Tables — the read path the platform serves positions and balances from. Both modules are taught as design; neither has a rung yet, and so neither claims a latency, throughput, or posting number. The first such number lands when the first rung's harness runs (docs/trading/trading.roadmap.md, TRD.3 and TRD.5).

The two-store decision is argued, not assumed.

The design splits the difference along the settlement boundary. The Postgres canon the ledger posts to is the substrate of /redis-patterns and /elixir; the stream lanes the unsettled log moves to are /echomq. The module returns to the capstone at B8.2 · The Log and the Ledger; the strategies that emit intents over this same log are the subject of B8.3 · Strategies as Deciders, a PROPOSED forward module.