experiment is swapped or 2) the experiment is completely terminated.
In these case, lets put explicit swapout/destroy events into
testbed_stats so that the record is not confused by experiments that
appear to start when they are still running. This really throws off
the summary stats web page!