why some time vm connector in polling stops processing all the messages in queue?

I found that in cloud hub some time the queues are remain unprocessed.I am polling for each 2 min every time i get around 500 records from DB and process them,but sometime when my polling starts it took all the records from DB but after processing few records it stops processing the other records.I am using one way vm and i have enabled persistent queues also.On what condition the queues will remain unprocessed.i dint find any exception also although i am handling the exceptions.

Thanks