hmmm. odd. it shouldn't be trying to sync it anyway, raid should have marked the drive as failed by now. if 'mdadm --remove' doesn't work, you could try 'mdadm --fail'
I ran into this exact issue over the weekend. A raid array was in the middle of rebuild, and the SATA drives was tossing errors but the rebuild was not failing.
This cmd stopped the rebuild immediately and marked the array member as failed:
mdadm -f /dev/mdX /dev/sdb7 Where X is your md device and /dev/sdb7 (in my case) was the member.