The following workaround is unverified, as the problem has not been reproduced manually. However, running this command for the affected node should solve the problem.
hadbm restartnode --level=clear nodeno dbname |
Note that all devices for the node will be reinitialized. You may have to stop the node before reinitializing it.