I am going to give you an example of what I worked on for 3 months before we figure it out, and you can see if it sounds like your issue.
I had jobs that would start and say they were writing.
I may or may not still see the tapes in the drives.
I could not just cancle the job.
I would have to kill the bpsched's ( yes many running when they should not be more then one)
I was on a windows master with unix media servers
I would have to stop services, remove some files
then restart services or reboot.
I saw this EVERY day.
turns out that the write cache on the san drives was turned off. this caused the backups to go so slow that they could not even figure out how to die.
Once the san people turned the write cache back on it worked fine.