The JobScheduler is configured with need_db = yes in the ./config/factory.ini.
Now the database connection get lost.
If the JobScheduler doesn't have a cluster option then it waits (endless) until it gets the database.
All running tasks are still opened and will be resumed when the database is available.
That works properly.
If the JobScheduler has cluster options like -exclusive or -distibuted-orders then after ca. 2 minutes the current process shuts down and a new JobScheduler process starts. This new process waits for the database.
In this case all running task are terminated and the task log could not be written in the database.
The JobScheduler with cluster options should have the same behaviour then without cluster options.