There is performance regressions in Controls.TableView-XmasTree appeared in build fx2.2-graphics-scrum-324.
The results for #323 and #324 are:
 WinXP-mid: 21.5 => 18.5, -15% (-3.25 fps)
 Win7-high: 31.0 => 24.5, -21% (-6.5 fps)

I can easy reproduce this regression. Steps to run the benchmark are:
> cd JFX_WS/tests/performance/Controls
> ant
> java -Xmx512m -Djavafx.animation.fullspeed=true
       -cp "JFX_HOME/rt/lib/jfxrt.jar;./dist/Controls.jar;../FXBenchmark/dist/FXBenchmark.jar;../../../import/benchmarks-2.1.1/benchmarks-2.1.1.jar"
  jrockit.bm.Main controls.bm.XmasTreeTableViewBenchmark -i 2 -wt 10 -tr 20
  -win-width 1280 -win-height 1024 -col-width 25 -row-height 10 -rows 90 -columns 50 -batch-size 10


According to hudson fx2.2-graphics-scrum-324 contains the following changes:
    1. RT-21428 some small glitches with Remote WebKit Debugging
    2. merge
    3. RT-16011 Need mechanism for PG nodes to know when they are no longer part of a scene graph
    4. Automated merge with ssh://jfxsrc.us.oracle.com//javafx/2.2/scrum/graphics/jfx/rt
    5. [DOC-ONLY] Replace setRepeatCount() with setCylceCount() in documentation [RT-21604]
    6. merge
    7. RT-16011 Need mechanism for PG nodes to know when they are no longer part of a scene graph


Perhaps the regression is caused by RT-16011.

Will try to collect profile data and provide more details.