This applies in principle only to those devices that have at least 2 projects attached with work in the Task buffer (cache), for instance World Community Grid and Project X!

The client scheduler is designed to allow each distributed computing project attached to a device their allocated amount of processing time. When the "Switch between applications every XX minutes" is set to e.g. the default of 60 minutes, the client waits for a job to reach the first checkpoint after the 60 minutes has passed before doing so. Occasionally that happens to coincide with the last checkpoint in a job. At that time a task will show a rounded 100% completion with "Waiting to Run", but in fact has not done the final verification and integrity check. Since the client just manages jobs and does not know the internal functioning of the science applications, it will just act on it's own scheduling routine of "checkpoint done, times up, it's another project's turn to get its day-part segment".

Action: None! Just let it sit and the client will eventually switch back to finish the "seemingly" completed task.

The complete FAQ index is found here