Sharding & scatter scan

A deployment runs K independent VSR shard groups behind a router with a rendezvous key→shard mapping. A single-shard transaction stays on its shard's own VSR group (serializable, fast path). Cross-shard Op::Txn is deterministic (Calvin-style) — slices durably totally ordered by a sequencer group, then each shard applies its slice in that order via a decide → commit — no 2PC, no coordinator-failure hole.

Cross-shard reads (SP-A) — Select / QueryRows / SelectFields / SelectSorted automatically scatter across every shard via scatter_scan. Unordered scatter = shard-id-deterministic concatenation. Sorted scatter = BinaryHeap k-way merge of already-sorted per-shard streams. K-invariance is locked across K ∈ {1, 2, 4, 8, 16} by an 85-seed property sweep — with unique sort values, merged output is byte-identical to the K=1 baseline.

Full reference: Architecture → Sharding & cross-shard transactions and Architecture → Cross-shard reads (SP-A).