Very possibly that the jobs are crashing the Grid Server \ Grid Nodes.
You can check the CCA UI --> Administration --> Grid Information --> Grid Jobs --> see if any exceptions are listed in the Exception Type column. If so, view those, there might be errors that are crashing the nodes such as Java Heap space or maybe a near-infinite scan that is holding up the job
Also, check what blueprints you are discovering, as at least one of those BPs may be effecting the discovery. If you have any custom BPs there, I would remove them and test with the OOTB BPs first, if that works, add one Custom BP at a time to see if you can find a culprit
If no Custom BPs exist, then check if any of the blueprints are discovering a "heavy" application, like Exchange, or a Domain Server where a vast amount of information has a potential of being discovered
Another test (which you may want to mix in somewhere there) is if you do have Grid Nodes, stop them. Let the discovery only use the CCA Server's Grid Server. This will impact the server's performance a bit if you run many big MPs, so I would do one at a time. It's a useful troubleshooting technique
If none of the above helps or don't make enough process, I would recommend opening a support ticket and possible doing a remote session with one of the support engineers