Problem
Background loading, viewport culling, lazy image decode, and bounded caches exist, but there are no recorded hardware baselines or regression thresholds for large workspaces and pages.
Acceptance criteria
- Repeatable fixtures measure cold parsing/indexing, incremental replacement, and warm search.
- Viewer measurements cover startup, page switching, pan/zoom, long-page scrolling, and image-heavy pages.
- Multi-notebook memory use and task scheduling are measured.
- Hardware, software stack, corpus shape, latency percentiles, and peak memory are recorded.
- Stable microbenchmarks run in CI; environment-sensitive GUI measurements have a documented manual process.
- Optimization work is justified by measured bottlenecks.
Problem
Background loading, viewport culling, lazy image decode, and bounded caches exist, but there are no recorded hardware baselines or regression thresholds for large workspaces and pages.
Acceptance criteria