Cypher MATCH after WITH carried scalar columns rejects reentry unless the prefix MATCH seed is a single node #1004

Summary

Direct Cypher lowering currently rejects a benchmark-relevant MATCH ... WITH ... MATCH ... continuation when the WITH carries scalar columns and the trailing MATCH seed is not treated as a single-node prefix seed.

Current measured error:

Cypher MATCH after WITH carried scalar columns currently require a single-node prefix MATCH seed

Benchmark impact

This is the next blocker exposed by interactive-short-2 / expand-order-limit after connected-pattern variable-length lowering moves past #973.

Measured in the benchmark investigation captured in:

• /home/lmeyerov/Work/pyg-bench/plans/benchmark-roadmap/colleague-note-cypher-multihop-connected.md
• /home/lmeyerov/Work/pyg-bench/plans/benchmark-roadmap/plan.md

Why this is separate

This is not the connected-multihop lane itself. It is the downstream reentry / carried-scalar continuation blocker surfaced once that lane stops failfasting earlier.

It is related to:

• #999 broader multi-stage MATCH/WITH/MATCH reentry
• #1001 wrong rows in two-MATCH reentry with connected varlen
• #981 broader multi-binding row/projection limits

But this issue tracks the narrower benchmark-facing boundary first.

Expected direction

Support the narrow carried-scalar WITH -> MATCH continuation shape without adding non-vectorized row-loop fallback, or fail with a narrower explicit contract if investigation proves it is inseparable from the broader row-carrier architecture.