Recent benchmarks (#9199) identified a significant performance bottleneck in the Grid component on Desktop viewports (1920x1080).

• Mobile (375x667): 60 FPS
• Laptop (1366x768): 54 FPS
• Desktop (1920x1080): 30 FPS

The sharp drop to 30 FPS on Desktop, despite only a ~12% increase in active cells compared to Laptop, suggests a non-linear scaling issue or a specific CSS/Layout trigger that is disproportionately expensive at this scale.

Since general optimizations (throttling, buffer reduction) failed to improve the Desktop framerate, we need to isolate the specific rendering cost.

Strategy: Column Ablation
We need to systematically remove specific column types from the DevIndex grid and re-run the benchmark to identify if a particular component style is the culprit.

Tasks:

1. Modify Benchmark: Update test/playwright/e2e/GridScrollBenchmark.spec.mjs to support an excludeColumn parameter (or create variants).
2. Test Iterations: Run the benchmark on Desktop (1920x1080) for each scenario:
   • Baseline: All columns active (Control).
   • No Images: Exclude GitHubOrgs, CountryFlag, and IconLink (Avatar/Flag/Icon images).
   • No Sparklines: Already tested (30 -> 33 FPS, minor impact), but worth re-verifying in this suite.
   • No Custom Components: Exclude Heuristics.
   • Text Only: Hide all Year/Number columns (reduce text node count).
3. Analyze: Compare the FPS and Main Thread Trace activity.
   • If removing CountryFlag jumps FPS to 50+, investigate its shadow/border/SVG styles.
   • If removing GitHubOrgs jumps FPS, investigate border-radius or image decoding costs.
   • If "Text Only" is still slow, the issue is the raw layout cost of the CSS Grid/Flexbox structure itself.

Outcome: Identify the specific CSS/DOM feature causing the 30 FPS cap on large screens.