I'm still not keen on working around a Tk crash by modifying IDLE especially when we know exactly which Tk change set caused the regression.  On the other hand, I sympathize with Kevin regarding the complexity of trying to track down and fix things involving event loops spanning Cocoa, Tk, Python, and IDLE and I do appreciate the effort you've put into debugging this and other Tk problems.  Thank you, Kevin.

So, if we're going to take the expedient approach, we need to come up with proper patches for 2.7, 3.2, 3.3, and default and then we need to test them for regressions with our other major supported platforms: Windows, some set of X11, OS X Carbon Tk 8.4, as well as the Cocoa Tk 8.5.  I can test the latter two.  I'd appreciate it if others could test the other platforms.  Roger, do you feel up to turning your suggestions into patches?  Once we're comfortable that the workaround works and doesn't cause other regressions, I can bug the ActiveState folks to push out a new release of ActiveTcl with fixes for some other problems and these fixes can go out in the next set of Python maintenance releases that should be coming up in the near future.