[imakris]

A preview that matches its main series currently copies the main LOD scale vector even though the planner accepts it by const reference and the main vector remains alive throughout both calls.

Changes

• Pass main_scales by const reference for matching previews.
• Build a separate preview scale vector only when a valid preview uses a different configuration.

Validation

These are retained results for the unchanged candidate; task-start checks confirm its base files still match GitHub master. An independent source review found no blocker.

• Twelve source-extracted scale-selection cases preserved selected values and source-read counts across matching/different/invalid previews and 1, 8 and 128 levels.
• The matching valid-preview case removed exactly one vector allocation; other cases retained their allocation counts.
• git diff --check passed.

The configured public PR workflows passed as recorded below; existing preview/planner coverage remains authoritative.

Evidence limits

• The local Qt 6.12 development SDK is unavailable (Qt6_FOUND=0); native build/test evidence comes from the published CI runs below. No local GPU or application performance run is claimed.
• No end-to-end Logonomic performance measurement was performed. These draft changes are candidates for native validation and representative performance qualification before production adoption.
• Pure-C++ source-extraction probes use stubs only for surrounding unavailable framework objects; they prove their named helper boundaries, not Qt integration.
• Savings are deliberately small for sources with only one LOD level.

Published native CI

Automatic pull-request workflows passed for published head 4f6aaa958137c09377c8ff968547a254d6cd1c50:

• CI Linux: passed.
• CI macOS: passed.
• CI FreeBSD: passed.
• CI Windows: passed.

These workflow results supplement the retained helper probes above. Representative application/rendering performance qualification remains outstanding.