There is following regression in build 8.0-graphics-scrum-h276:
 NodeMemory.TableColumn-rendering-on: -33%, 1,620 objects => 1,080 objects

According to MemoryAnalyzer the number of TableColumnHeader objects is created in twice bigger than
the number of TableColumn objects. I.e. for each TableColumn object 2 TableColumnHeader objects are created.

It looks like the regression was actually introduced in build 8.0-graphics-scrum-h275 (which was failed).
According to hudson it contains:
    1. merge — Martin Sladecek / detail
    2. RT-25363 Try to optimize memory footprint and performance of ListChangeBuilder — Martin Sladecek / detail
    3. merge — Martin Sladecek / detail
    4. RT-9383 Add proper constructors & factory methods to event classes, remove impl — Martin Sladecek / detail
    5. merge — Martin Sladecek / detail
    6. RT-9383 Add proper constructors & factory methods to event classes, remove impl — Martin Sladecek / detail


The regression is easy reproduced. Steps to run the benchmark:
 > cd JFX_WS/tests/performance/NodeMemory/
 > ant
 > java -client -Xms64m -Xmx64m
       -cp "JFX_HOME/rt/lib/jfxrt.jar;./dist/NodeMemory.jar;../FXBenchmark/dist/FXBenchmark.jar;../../../import/benchmarks-2.1.1/benchmarks-2.1.1.jar"
       jrockit.bm.Main nodememory.bm.TableColumnBenchmark -i 1 -tr 120 -wt 0 -render true -allocCnt 10



This memory leak is most probably is the reason of Controls.Buttons regression which
also appeared in build 8.0-graphics-scrum-h276.