It is very difficult to replicate these bugs, you can try submitting a lot of collections setting short expire and matchmaking periods (i.e. ExpiryPeriod and MatchRetryPeriod) in the configuration file and then restarting some times the workload manager on the WMS to trigger the recovery mechanism.
Submit a collection with at least a node that doesn't match (e.g. put requirements=false)
After the match making of the collection, the node should be put in state Waiting with reason: "BrokerHelper: no compatible resources"
Check if the request of the collection is in the directory $GLITE_LOCATION_VAR/workload_manager/jobdir/old (i.e. you should find a file like 20081218T092023.381155_3086493376)
Manually restart the wm on the WMS (i.e. $GLITE_WMS_LOCATION/etc/init.d/glite-wms-wm restart)
In a while the recovery process should re-match the node and the request should be put again in the $GLITE_LOCATION_VAR/workload_manager/jobdir/old directory
Check that after the node request will expire (i.e. after ExpiryPeriod seconds), at the next matching session the node should be aborted (with reason: "request expired") and the file should be removed from the $GLITE_LOCATION_VAR/workload_manager/jobdir/old directory