The documentation is not clear regarding what triggers a NiFi flow execution to be considered as a failed Kylo feed.
Based on my current understanding, a NiFi flow is considered to be failing, if the flow file transitions through at least one NiFi queue, whose name contains the string "FAILURE".
Without this information, it is relatively hard to figure out how to intentionally cause a Kylo feed to be displayed as FAILED. In addition to that, if a NiFi flow is designed so that it contains a queue with the name "FAILURE" during normal operation (not encouraged, but valid scenario), one will have some fun till he/she figures it out why the corresponding Kylo Feed keeps failing, while on the NiFi side, everything goes as expected.
The closest description I could find in the best practices section is relatively vague:
Kylo is configured to look for any activity along standard failure paths and trigger alerts in Ops Mgr.