Virtual Agent triggers fallback topic twice when neither the NLU or the keyword search have matchingDescriptionSetting glide.ml.async.http.max.outstanding.requests = 0 and glide.nlu.async.ml.predict.batch = true can cause a service_error, which can lead to duplicate "I cannot find your topic" fallback topic executions. VA should not trigger the fallback topic twice when neither the NLU or the keyword search have matching (with Async NLU configured).Steps to Reproduce Navigate to open_nlu_driver.listEnable ServiceNow NLU by setting 'Asynchronous Prediction Supported' to trueAdd the following to the sys_properties.list:glide.mlpredictor.option.nlu.disableNER = trueglide.mlpredictor.option.nlu.turnOffInferenceParsing = {'modelPurposes':['virtual_agent']}glide.ml.async.http.max.outstanding.requests = 0glide.nlu.async.ml.predict.batch = trueEnable the NLU by navigating to Virtual Agent > General SettingsCreate a NLU model 'Test Model' and an intent 'Test Intent' with the utterance 'nothing'SaveTrain and publish the modelCreate a topic 'TestTopic' in Virtual Agent > DesignerAssign the 'Test Model' and 'Test Intent' to itClick the 'Test Topic'Enter 'blah' in the test window with the VA simulation Immediately after the Show me everything buttons appear, the VA, which should not find any match, repeats again 'I am sorry but I didn't understand ....' and shows the Show me everything button again.WorkaroundThis problem has been fixed. There is no workaround available. The system property 'com.glide.cs.nlu.keywords.enabled' has been changed to be set as true by default as part of the fix. If you are able to upgrade, review the Fixed In section to determine the latest version with a permanent fix your instance can be upgraded to.Related Problem: PRB1482906