Ralf (TPCBF), In each slot there is a file like this...

[root@danda slots]# cat ./0/boinc_task_state.xml
<active_task>
    <project_master_url>http://www.worldcommunitygrid.org/</project_master_url>
    <result_name>SCC1_0003264_RelA-A_3778_0</result_name>
    <checkpoint_cpu_time>2292.900000</checkpoint_cpu_time>
    <checkpoint_elapsed_time>2308.332923</checkpoint_elapsed_time>
    <fraction_done>0.649393</fraction_done>
    <peak_working_set_size>45703168</peak_working_set_size>
    <peak_swap_size>122249216</peak_swap_size>
    <peak_disk_usage>713790</peak_disk_usage>
</active_task>

This gives you the task and progress (SCC1_0003264_RelA-A_3778_0 and 0.649393 in this example)...
Shouldn't be too hard to create a script running periodically on each remote system checking that file in each slot for too little or no progress in <fraction_done>, for the same task <result_name>, and using the relevant boinc command to abort it if required ...