The WCG server doesn't have any clue about tasks getting or being stuck, because the BOINC client (also) doesn't know. The BOINC client is the link between the tasks on your computer and the WCG server.

The BOINC client isn't aware of stuck tasks, because it doesn't know about any definition of stuck. However, it does know about a time limit. When that limit is exceeded, your 'stuck' task will be aborted.

In the past, I've had GPU tasks that had a time limit of 1.68 hours (100 minutes) with an expected duration of only 3 minutes, while they needed more than 100 minutes to run. Sure enough they were aborted by the BOINC client after 100 minutes of runtime. That's the BOINC client saying "Your task is taking too long, I'm aborting it." It doesn't say "Your task is stuck, I'm aborting it." However, it's the BOINC client's way of saying "Your task is stuck, so I'm aborting it" (if you want to look at it that way ).



Then you need a computer program running in the background to check every now and then if there is a task that exceeds your limits, meeting your definition of being stuck.

EDIT: added a simple program:

Say you would run this script, letting the process sleep every 5 minutes till the next check, then letting it detect if there is an OPNG-task running for at least 10 minutes, and if there is, abort it. With echoed comments. For your pleasure. Use at your own risk.

while sleep 300; do
  printf "[%s] " "`date`";
  wcgresults -HNrB | grep OPNG |
    while read elapsed name; do
      case $elapsed in
        (0:0[0-9]:[0-5][0-9])
  	  echo Steady as she goes, $name, elapsed time = $elapsed
  	;;
        (*)
  	  echo Whoa! Task $name needs attention, elapsed time = $elapsed, aborting it; 
  	  boinccmd --task http://www.worldcommunitygrid.org/ $name abort
  	;;
      esac;
    done;
done