Comment 7 for bug 780266

I have found a way to reduce CPU load without entirely stopping the udev queue.

Create /etc/udev/rules.d/my.rules file, and put the following line into it:
KERNEL=="sr1", ACTION=="change", WAIT_FOR="nothing"

This will introduce a 10 seconds delay when processing 'change' events from /dev/sr1. As a result, CPU load returns to normal. I did not notice any negative side effects so far. Other udev events seem to get processed normally.

This is still a work-around, not a solution. The solution would be to find out exactly what's wrong with the existing udev rules, and fix the problem there.