019 store batch insert fused cascade

○ Planned This feature is planned but not yet implemented.

IDEA PIN (residue-tier, no spelling yet): batch insert + full cascade fusion — T2's acyclic comptime cascade DAG topo-sorted and fused into one straight-line body per write site, batches applied column-wise over SoA (SIMD), batch = the transaction envelope. See residue.md.

Supporting Files

residue.md

Test Configuration

NEEDS_RULING LANGUAGES: zig cs

Awaiting ruling:

Does a batch insert fire per-row watches N times after the fused apply, or once with a plural payload? The pin's own lean, recorded in residue.md: a plural arm (`! inserted rows` — spelling provisional), with per-row as the degenerate case, because that is the shape that keeps the fused fast path honest. residue.md calls it "Genuine semantics call, Lars's", which is why it sits here rather than in a design doc. The batch-insert SURFACE spelling is uninvented and stays that way — do not write it into a test before the semantics are ruled. Adjacent and separable: the delta-algebra soundness question the adversary round flagged for MIN/MAX-under-removal applies verbatim to a fold edge under vectorization. That is a correctness question about the planner's edge classification, not this ruling. Blocks: this pin (residue-tier, no spelling yet), and the fused-cascade path it stands for. Sibling question, same shape, different subject: 690_017 asks it about DI arms.