Tom, thanks for the detailed analysis.  It seems pretty clear that this is not a Python problem and there's really nothing we can do about it.  It would be great if you would be willing to open an issue on the Tcl/Tk project's issue tracker so that they are aware of it and perhaps can investigate it now that there might be a way to reliably reproduce the crash.  I'm going to close this issue under the assumption it is a Tk problem.  If it should prove otherwise, feel free to re-open.

http://core.tcl.tk/tcl/reportlist