When a client has been off for longer, or scheduling projections have led to very long estimated return times, further out then the deadline of the chosen projects, work will not be sent to these clients, sometimes up to a point where cached work has been whole or partly completed. When the condition arises, a signal will be recorded in the BOINC message log, similar to the sample below:

22-01-2010 14:44:37 World Community Grid Message from server: (won't
finish in time) Computer on 41.1% of time, BOINC on 86.7% of that

Whilst BOINC from later 6.10 versions is much better at handling these conditions [to include ignoring periods when the client has been off for longer then 10 days], it could still occur. Restoring the client to a condition of work fetch can be achieved by leaving the computer on for longer so BOINC thinks there's more time to work on the science and eventually some will be retrieved, but that's not very satisfying, let alone an energy friendly way. There is a workaround to quickly fix the problem:

• Only if there is no work in progress, likely not, open the BOINC Manager Advanced view, go to the projects tab, select WCG and push the Reset Project button. This restores default for the WCG specific Duration Correction Factor aka DCF. Also visit the Activity Menu and select the Run Always option. Wait a bit and if tasks are fetched, skip all steps below.
  
• Find the exact location of the BOINC data directory. The path is
  found in the message log near the beginning, for Windows for instance
  
  21/01/2010 20:02:57 Data directory: C:\ProgramData\BOINC
  
  
• Exit BOINC completely. Verify if need be in the Process/Task Manager, with admin privilege if installed as service.
  
• Go with the operating system's file navigator to the data_dir of BOINC as found under point 2 above.
  
• Make a backup copy of the client_state.xml file for safety, though there is a periodic backup already called client_state_prev.xml
  
• Open the client_state.xml with an ASCII file editor, not one that can produce hidden characters, and find the <time_stats> section. Delete this line and all after up to inclusive </time_stats>.
  
• Save the file making sure it's name has remained exactly the same. If need be select the ANSI file type encoding or change the save as text document to
• all documents[]*, so nothing in invisible characters gets added and the file name/extension remains unchanged.
  
• Start BOINC again in user or service/daemon mode.

This should rebuild the time keeping statistics back to their standard settings as were it a new install. Within a brief period tasks should be fetched again for the projects that are allowed to.

Related topics:

Reporting x completed tasks, not requesting new tasks
[sched_op_debug] CPU work request: 0.00 seconds; 0.00 idle CPUs