The customer has supplied a test case that reproduces the error case reliably: for_red_hat_use_cases_sanitized.zip

Note: when trying the fail case your ESB instance will lock up, requiring a kill -9.

How to reproduce. Follow README.TXT to run app from for_red_hat_use_cases_sanitized.zip

and invoke JMX (for example from jconsole):

com.my.camel.beans:obj=JMXCamelInvoker
invoke operation: invokeCamel with parameters "direct:dummy", "a", "<a/>
you will get: OK response: <a/>

invoke operation: invokeCamel with parameters "direct:goToNotExistingJBIEndpoint", "a", "<a/>
you will get: OK response: <error>Error: java.lang.IllegalStateException: No endpoint, service or interface name specified for routing occurred in endpoint:https://my.namespace.com/GhostEndpoint/my during processing</error>
what is correct behavior

BUT

invoke JMX
com.my.jbi.beans:obj=JMXJBIInvoker
invoke operation: sendAction with parameter "OTA_PingCamel" after 10 seconds (a timeout ) you will get: "Response didn't come" and the logs are constantly growing with:

15:21:53,062 | INFO | rovider-thread-5 | camel | 119 - com.myproject.camel-endpoints - 1.0.0.SNAPSHOT | OTA_PingCamel route processing has started
15:21:53,062 | INFO | rovider-thread-4 | camel | 119 - com.myproject.camel-endpoints - 1.0.0.SNAPSHOT | OTA_PingCamel route processing has started
15:21:53,062 | INFO | rovider-thread-6 | camel | 119 - com.myproject.camel-endpoints - 1.0.0.SNAPSHOT | OTA_PingCamel route processing has started
15:21:53,063 | INFO | rovider-thread-7 | camel | 119 - com.myproject.camel-endpoints - 1.0.0.SNAPSHOT | OTA_PingCamel route processing has started

In conclusion:

So having flow:
Camel -> not exisiting JBI endpoint

Invoking directly that Camel flow - everything works.
Invoking Camel from JBI which point to not existing Camel flow - causes that Camel rerun the same flow endlessly.

Enjoy!

Forgot to say that it ends with: FuseESB:karaf@root> Exception in thread "RMI TCP Connection(idle)" java.lang.OutOfMemoryError: GC overhead limit exceeded