When you have a processor in a flow which fails on some exception which is not propagated to any of the processor's relationship (e.g. processor doesn't have a failure relationship or it doesn't expect exceptions and doesn't handle them; processor which meets both conditions is for example UpdateAttribute processor) such processor is not marked in the feed job's details as failed and such job runs indefinitely.
Also this failing processor generates a lot of messages which are sent to message queue and can make it stuck (see attached screenshot) which then prevents Kylo for reporting job status for other feeds.
Stacktrace for exception which occurred in UpdateAttribute processor can be found in file attribute_long_exception.txt which after few repeated runs changes to an exception in file attribute_long_continuous_exception.txt.
Exception is caused by having a long string in an attribute's value (in this case metadata.table.fieldPoliciesJson). More details can be found in NiFi's JIRA ticket https://issues.apache.org/jira/browse/NIFI-3389