PortletStateManager.clearState(PortletWindow) implementation is buggy and can lead to ConcurrentModificationException under stress.

Under stress, you can
run into the following scenaro which produces this stack trace.

java.util.ConcurrentModificationException
      at java.util.HashMap$HashIterator.nextEntry(HashMap.java:782)
      at java.util.HashMap$KeyIterator.next(HashMap.java:818)
      at org.jasig.portal.container.services.information.PortletStateManager.clearState(PortletStateManager.java:355)
      at
org.jasig.portal.channels.portlet.CPortletAdapter.receiveEvent(CPortletAdapt
er.java:345)
      at
org.jasig.portal.MultithreadedCharacterChannelAdapter.receiveEvent(Multithre
adedCharacterChannelAdapter.java:62)
      at
org.jasig.portal.ChannelManager.finishedSession(ChannelManager.java:269)

After examining the code, it looks like you'd get this exception when two
users end/begin the lifecycle of this channel at the same time (likely under
stress??), since they both have access to the same Map object at the same
time -- classic concurrency stuff. And once the remove operation throws
that exception, I believe the ChannelState object get left behind. It looks
like remove's and put's into the channelStateMap need to broken out into
synchronized methods.