[kvlonge]

Why

Snowflake exposes unaliased VALUES fields as COLUMN1, COLUMN2, and so on. The native schema validator reported those valid direct projections as unknown whenever a complete project schema was supplied.

Changes

• discard that native diagnostic only after AST evidence proves the exact unqualified direct projection and VALUES ordinal in the same SELECT scope
• retain unknown-column findings for out-of-range ordinals, other relations, joins, aliases, quoted identifiers, and other dialects
• add focused positive and negative schema-validation coverage

Verification

• focused schema-validation tests: 9 passed
• Ruff and Ty checks passed for touched files
• Fensu passed
• uv sync --all-extras
• make check-ci
• public-repository hygiene scan
• cache-free static compile of the downstream 974-model project with the affected analysis boundary enabled: 0 errors, 0 warnings