# Changelog

## Score

- CAI 51 → 53 (+1.3)
- Rubric changed (rubric-2026.10.5 → rubric-2026.11.2) — scores are not directly comparable.

## Lenses

- Code Health 64 → 64 (+0.0)
- Architecture 95 → 95 (+0.0)
- Maturity 61 → 65 (+4.1)
- Readiness 54 → 54 (+0.0)
- Security 42 → 44 (+2.1)
- Event Sourcing 100 → 100 (+0.0)
- Accessibility 56 → 57 (+0.8)
- Performance 91 → 91 (+0.0)

## Resolved (5)

- Ambiguous naming for sequence generation. `EventSequencer.sequence` takes an event and returns a u64 (likely the assigned sequence number), while `SequenceGenerator.next` returns a u64 (the next sequence number). It is unclear if `sequence` is a verb (assigning) or a noun (fetching). Furthermore, `EventSequencer` has `current_seq()` while `SequenceGenerator` has `peek()`. `peek` usually implies looking at the *next* value without consuming, whereas `current_seq` implies the *last* or *current* value. This creates confusion about whether the API exposes the next available ID or the last assigned ID.
- Inconsistent error handling for empty/failed operations. `SpscRing.try_pop` returns `T` directly (implying it never fails or panics on empty, or `T` is an Option/Result wrapper not shown but implied by 'try'). `SpscRing.try_push` returns `Result`. `EventSequencer.drain` and `ReplayPlayer.next` return `Event` directly. If `try_pop` can fail, it should return `Result<T>` or `Option<T>` to match `try_push`. If `drain`/`next` can fail (e.g., empty), they should also return `Result` or `Option`. The mix of direct returns and Result types for similar 'get next item' operations is inconsistent.
- Inconsistent iterator/consumer semantics for sequential data. `EventSequencer` uses `drain` (implying removal/consumption) and `peek` (read-only). `ReplayPlayer` uses `next` (consumption) and `peek` (read-only). The naming `drain` vs `next` is confusing when both serve the role of advancing the cursor and returning the current item.
- Inconsistent return types for collection queries. `open_orders` returns `LiveOrder` (singular) but implies a collection. `open_order_ids` returns `OrderId` (singular). `stale_order_ids` returns `OrderId` (singular). These methods likely return collections (Vec/HashMap) but are named and typed as if returning a single item. This is a critical API design flaw.
- Inconsistent return types for similar queries. `OrderBook.best_bid()` returns a `Level` (price + qty), while `HotPathEngine.best_bid()` returns a `Price`. While the domains are distinct (OrderBook vs HotPathEngine), the semantic intent is the same (get best bid). Users must know which type to use based on the engine. If `HotPathEngine` is a simplified view, it should be documented, but the inconsistency in 'best' accessors is notable.

## New (2)

- Medium advisory (unmaintained): RUSTSEC-2026-0441 (Cargo.lock)
- Medium advisory (unsound): RUSTSEC-2026-0386 (Cargo.lock)
