My first thought was, no way would I do this. If the Agent is running on both then it you have no way of know which one will run a job, I think there is also a question about concurrency, what happens if BOTH start a job. I can't think of anything good to come from this..

I think we need to know what errors you are getting.

Worst case is uninstall that SQL instance (yielding a single node cluster), and then re-install it. You lose the second node for THAT instance for a bit but then you get it back..

CEWII