Cohort Segmentation

Use this when "large document" or "big surface" is treated as one bucket.

Rule

Segment by size and complexity before choosing tactics.

Baseline Slate cohorts:

Cohort	Examples	Default stance
normal	0-500 blocks, low decorations	optimize the repeated unit
medium	500-2000 blocks	DOM-present, strict budgets
large	2000-10000 blocks	DOM-present grouping, staged work, native behavior guarded
stress	10000-50000 blocks	explicit degradation candidates
pathological	custom renderers, comments, annotations, nested hidden ranges	complexity-tagged, not hidden inside block count

Complexity Tags

• custom leaf/text/element renderer
• decorations per block
• annotations/comments per block
• hidden boundary count and depth
• inline voids, voids, tables
• collaboration activity
• selection span length
• shell/DOM-present/off/staged mode
• mobile/IME/browser

Output

Every perf claim names the cohort it covers. No "fast for large docs" without a size and complexity tag.