Troubleshooting this here with limited information is going to prove difficult.
Does the downstream processor of the connection with the "stuck" FlowFiles have...
1. Configured for all nodes or primary node Execution? If primary node only, are the queued FlowFiles on the primary node?
2. Does processor show it has active threads? (small number in upper right corner of processor) . If so, get some thread dumps from your NiFi and inspect if the thread is progressing or hung.
3. Is back pressure being applied on any of the outbound connections?
4. Is processor getting threads to do work? What is CPU load on your system and configured "Max Timer Driven Thread Count?" It may be set to low causing a thread starved situation for your dataflow.
Additionally, if you try to perform a "list queue" on the connection of issue, what result do you get?
If you have a support contract with Cloudera, you can open a support case and they can help you to a solution.