It looks like our infinitely expanding pages are back when you run with a standalone rice server on a different host. I had seen this before but got it again just before the break. It happens because of the cross-site-scripting limitations. The code which checks the external frame's size can't get the real number and the browser then returns zero. This causes the code to attempt to make the frame wider and so on...
It happens when using the portal screens from KFS and then accessing a Rice screen (like the KIM lookups or the parameter document)
I observed this in both Firefox and Safari.
It seems that we need some way to detect when we are receiving a zero because of a scripting block like this and shut down the resizing code.