This issue may or may not be a bug. The suspicion is that it might impact performance:

Run the attached program when 8u20 JAB is activated: You will see this output in cmd:

java -jar ExampleAccessBridge.jar
....
JTable Renderer Thread AWT-EventQueue-0
JTable Renderer Thread AWT-EventQueue-0
JTable Renderer Thread AWT-EventQueue-0
JTable Renderer Thread Thread-3
JTable Renderer Thread AWT-EventQueue-0
JTable Renderer Thread AWT-EventQueue-0
JTable Renderer Thread AWT-EventQueue-0
JTable Renderer Thread AWT-EventQueue-0
JTable Renderer Thread AWT-EventQueue-0
JTable Renderer Thread AWT-EventQueue-0
JTable Renderer Thread AWT-EventQueue-0
JTable Renderer Thread AWT-EventQueue-0
JTable Renderer Thread AWT-EventQueue-0
JTable Renderer Thread AWT-EventQueue-0
JTable Renderer Thread AWT-EventQueue-0
JTable Renderer Thread AWT-EventQueue-0
JTable Renderer Thread AWT-EventQueue-0
JTable Renderer Thread AWT-EventQueue-0
JTable Renderer Thread AWT-EventQueue-0
JTable Renderer Thread AWT-EventQueue-0
JTable Renderer Thread AWT-EventQueue-0
JTable Renderer Thread Thread-3
JTable Renderer Thread AWT-EventQueue-0
JTable Renderer Thread AWT-EventQueue-0
.....


Thread-3 is the accessibility thread. The output comes from a TableCellRenderer (and also
JList-Renderer) . See MultiLineRendererWorkAround.java which implements a
TableCellRenderer.

So the question is: Why does the accessibility thread create GUI components outside of
the event dispatch thread (EDT). Is this intention? Is this a design flaw? Could it affect performance?

The issue can be seen with Java 6 and Java 8u20 (not tested with Java 7)

There is no fix required as there is no functional impact. Right now this is just an observation which is confusing. If this is by design then feel free to close.