On the Tasks tab, down the left, there are buttons to show ALL tasks and ACTIVE tasks

Always retreat while shooting ;>)

That checkpoint thing has been proposed before. One view is, that the client does not really know which job checkpoints when... it only tells tasks how often they may, if they want to. Suppose you have a mix of many sciences on an 8 core. One checkpoints every few minutes, the other every 10, and ol CEP2 does one in 10-300 minutes. If CEP2 were to take 300 minutes, you want the other cores to idle 290? Let me vote No. Very maybe, if one could pick a specific job and tell the client to suspend/permit power-down when it checkpoints. You sell the case to the developers and see if Dr. A. can be hooked. The developers mail list is rather 'we take no prisoners'. Then, do not stick to 6.2.19 and expect others to do all the testing, for it sure will come with bugs and unexpected angles.

There's various options that can be passed with boinccmd... offhand, one is to finish tasks, then switch off. As it stands, the future will bring likely less buttons, not more... you proved the case why that is considered. The "show active" is one to go, once they root out the cause of latency when many tasks are buffered.

[OT] @Sekerob:
"That checkpoint thing has been proposed before. One view is, that the client does not really know which job checkpoints when ..."
If checkpointdebug is on, a message appears in the log each time a task checkpoints. Who puts those messages there? The task, the client, or the manager? Probably not the task, as it would have to negotiate write-access to the log file. So, the BOINC system probably knows when each task checkpoints.
"you want the other cores to idle 290? Let me vote No"
Sorry, I didn't fully explain. My proposed "Suspend at next checkpoint" button in the Tasks tab would work the same way as the current plain ordinary "Suspend" button, except the selected tasks would each wait until it checkpointed before BOINC would suspend it. While a task had a suspend pending, its status would show as -um- "Suspend pending". When the task went into suspension, if there was another task able to run, that one would start, as per an ordinary suspend. So, if you wanted to shut down or reboot your system, you would first ensure that you had some tasks available to run, that checkpoint frequently. Then you would suspend all of your infrequent checkpointers (aka CEP2) that were not running. Next you would request your running CEP2s to suspend after each of them checkpoints. You then tie a piece of string around your finger to remind you to come back before the machine runs out of work, and you can leave for a while. When you come back, your running CEP2s may have suspended or finished, so you can shut down with minimum loss of crunching time, in most cases sooner than if you'd had to wait until all of your running CEP2s finished.

Alternatively, someone with well-oiled shell-script skills could probably knock up a script that polls the log file every few minutes and issues boinccmds to suspend the tasks when they checkpoint. Not having the ability in BOINC manager to copy text from the tasks tab would make it harder to supply the script with the names of the tasks that you want to suspend though.
[/OT]
"P.S. You will have pushed the button, as default is to show all FAIK."
Yeh, must have clicked "Show active tasks" without realising it.