This issue has been identified: it turned out to be a monitoring problem, and not a real file system issue on Oak.
Compute nodes are currently re-opening to new jobs, and everything should be operating normally again very shortly.